LABORATORY QUALITY CONTROL MANUAL
2nd Edition, 1972
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
Analytical Quality Control Program
Ada Facility
P. 0. Box 1198
Ada, Oklahoma 74820
-------
FOREWORD
Those who generate water quality data have a serious
responsibility to those who will use it. Such data often
become the basis for various action programs in water pol-
lution control. Among these are: (1) the construction and
operation of wastewater treatment works costing millions of
dollars; (2) the early detection of trends in water quality
degradation that, if allowed to go unchecked, could result
in the loss of beneficial water uses; and (3) court actions
that could result in the levying of heavy fines and other
penalties and even industrial shutdowns.
The significance of water quality data precludes any
thought of careless laboratory operation; however, even the
best staffed, equipped, and maintained laboratories need some
measure of product quality. Conscientious personnel and well
equipped laboratories are not enough.
The Environmental Protection Agency (EPA) is concerned
about laboratory quality and has initiated a program of improved
effort in that direction. This manual deals with two areas
of that program; statistical analytical quality control and
record keeping. Product quality control is an old technique
of the manufacturing industries. The statistics underlying
product quality control are a proven technique but their
application to routine laboratory production is new. This
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
SURVEILLANCE AND ANALYSIS DIVISION
REGION VI
ANALYTICAL QUALITY CONTROL PROGRAM
The Environmental Protection Agency (EPA) gathers water
quality data to determine compliance with water quality standards,
to provide information for planning of water resources development,
to determine the effectiveness of pollution abatement procedures,
and to assist in research activities. In a large measure, the
success of the pollution control program rests upon the reliability
of the information provided by the data collection activities.
To insure the reliability of physical, chemical, and biological
data, EPA's Division of Research has established the Analytical
Quality Control (AQC) Laboratory at 1014 Broadway, Cincinnati, Ohio.
The AQC program conducted by this Laboratory is designed to assure
the validity and, where necessary, the legal defensibility of all
water quality information collected by EPA.
The AQC Laboratory is responsible for:
Conducting analytical methods research, providing leader-
ship in the selection of laboratory procedures, conducting
a reference sample program for methods verification and
laboratory performance, and advising laboratories in the
development of internal qualitv control. In addition, the
-------
Laboratory develops and evaluates automatic water quality
monitoring instrumentation and assists EPA's ten Regions
in the procurement and installation of this type of equipment.
METHODS RESEARCH
Although analytical methods are available for most of the
routine measurements used in water pollution control, there is a
continuing need for improvement in sensitivity, precision, accuracy,
and speed. Development is required to take advantage of modern
instrumentation in the water laboratory. In microbiology, the use
of new bacterial indicators of pollution, including pathogens, creates
a need for rapid identification and counting procedures. Biological
collection methods need to be standardized to permit efficient
interchange of data. The AQC Laboratory devotes its research efforts
to the improvement of the routine "tools of the trade."
METHODS SELECTION
Assisted by Advisory Committees, the AQC Laboratory provides a
program for selecting the best procedures in water and waste analysis
from among those that are available. Through the publishing of EPA
methods manuals, updated regularly, the program insures the application
of uniform analytical methods in all laboratories of EPA. The val-
idity of the chosen procedures and the evaluation of analytical
performance are verified by reference sample studies involving
participation by regional, basin, and project laboratory staff
-------
personnel. The EPA methods manual is available to any organization
upon request.
INTRALABORATORY QUALITY CONTROL
To maintain a high level of performance in daily activities,
every analytical laboratory must provide a system of checks on the
accuracy of reported results. While this is a basic responsibility
of the analyst and his supervisor, the AQC Laboratory provides
guidance in the development of model programs which can be incor-
porated into the laboratory routine.
AQC REGIONAL COORDINATORS
The Administration-wide quality control program is carried out
through EPA Regional AQC Coordinators. The Coordinator, appointed
by the Regional Administrator, implements the program in his
regional laboratory and maintains relations with state and inter-
state pollution control agencies within the region to encourage
their use of EPA methods and active participation in the analytical
quality control effort. In addition, the Coordinator brings to
the attention of the AQC Laboratory the special needs of his region
in analytical methodology.
U. S. GEOLOGICAL SURVEY
Because water quality surveillance is a joint program between
EPA, the U. S. Geological Survev (USGS) and the states, the AQC
-------
Laboratory works closely with the USGS in securing uniform methods
in both agencies. Through regular interchange of procedural outlines
and joint participation in reference sample studies, the two agencies
seek to promote complete cooperation in water quality data acquisition.
PROFESSIONAL LIAISON
The Laboratory staff, along with other EPA scientists, actively
participates in the preparation of Standard Methods for the Examination
of Water and Wastewater (American Public Health Association) and in
subcommittee and task group activities of Committee D-19 of the
American Society for Testing and Materials. A senior member of the
AQC Laboratory staff is General Referee for Water, Subcommittee D,
of the Association of Official Analytical Chemists.
For further information write your Regional Coordinator or
Director
Analytical Quality Control Laboratory
Water Quality Office
Environmental Protection Agency
1014 Broadway
Cincinnati, Ohio 45202
-------
June 2S, 1971
-------
Regional Analytical Quality Control Coordinators
Francis T. Brezenski, AQC Coordinator
Environmental Protection Agency
Hudson-Delaware Basins Office
Edison, New Jersey 08817
Harold G. Brown, AQC Coordinator
Environmental Protection Agency
911 Walnut Street, Room 702
Kansas Citv, Missouri 64106
Charles Jones, Jr., AQC Coordinator
Environmental Protection Agency
1140 River Road
Charlottesville, Virginia 22901
Bobby G. Benefield, AQC Coordinator
Environmental Protection Agency
Ada Facility, P. 0. Box 1198
Ada, Oklahoma 74820
James H. Finger, AQC Coordinator
Environmental Protection Agency
Southeast Water Laboratory
College Station Road
Athens, Georgia 30601
Daniel F. Krawczyk, AQC Coordinator
Environmental Protection Agency
Pacific Northwest Water Laboratory
200 South 35th Street
Corvallis, Oregon 97330
LeRoy E. Scarce, AQC Coordinator
Environmental Protection Agency
1819 West Pershing Road
Chicago, Illinois 60605
Donald B. Mausshardt, AQC Coordinator
Environmental Protection Agency
Phelan Building, 760 Market Street
San Francisco, California 94102
Robert L. Booth, AQC Coordinator
Environmental Protection Agency
Analytical Quality Control Laboratory
1014 Broadway
Cincinnati, Ohio 45202
John Tilstra, AQC Coordinator
Environmental Protection Agency
Lincoln Tower Building, Suite 900
1860 Lincoln Street
Denver, Colorado 80203
-------
INTRODUCTION
The precision and accuracy of analytical data produced in the
laboratory can be detected by using quality control charts. These
control charts serve as "fingerprints" of a laboratory's operations.
In addition, the use of these charts enables a supervisor to validate
the data produced from a specific laboratory group for the analysis of
a specific parameter. These charts indicate when the laboratory is
operating normally or abnormally, thus pointing out when data generated
should be accepted, questioned, or rejected. In the same manner they
indicate when the laboratory is operating at optimum efficiency.
CONSTRUCTION OF CONTROL CHARTS
Two control charts are required to "fingerprint" the laboratory
operations for a given analytical procedure. These are referred to
as precision and accuracy control charts. Precision control charts
are constructed from duplicate sample analyses, whereas accuracy
control charts are constructed from spiked or standard sample analyses
data. A set of these control charts represents, and is restricted to,
a specific laboratory, group of analysts, analytical method, range of
concentration, and period of time. To construct the precision and
accuracy control charts it is recommended that at least 20 sets of
duplicate and 20 sets of spiked sample data from an in-control process
be used for the initial construction. The selection of in-control
-------
data can be made on a judgment basis.
It is necessary that the initial sets of data be obtained under
the following conditions:
1. Normal laboratory operations
2. Constant analyst or group of analysts
3. Consistent method
A. Narrow range of concentration of the
parameter analyzed.
Since the precision and accuracy of the analyses of many parameters
are proportional to the concentration of the parameter to be measured,
it nay be necessary to use several control charts in many different
ranges of concentrations for a given parameter.
The control charts are derived from three basic calculations:
1. Standard deviation (S,) of the differences
d
between duplicates or, in the case of spiked
or standard samples, between the known quantity
and the quantity obtained.
2. The upper control limit (UL)
3. The lover control limit (LL)
Prior to these calculations, two decisions must be made:
1. The a and 8 levels
2. The allowable variability levels
-------
By definition, a is the probability of judging the process to be
out of control, when in fact, it is in control. It is recommended that
a be chosen to lie between the boundaries of .05 and .15, that is, the
laboratory personnel are willing to stop the laboratory process some-
where between 5 and 15% of the time, judging it to be out of control,
when in fact, it is in control. If the cost of examining a- process to
determine the reason or reasons for being out of control is considerable,
then it may be desirable to choose a low a. Likewise, if the cost is
negligible, it may be desirable to choose a larger a value, and thus
stop the process more frequently.
On the other hand, g is defined as the probability of judging the
process to be in control when it is not. Again, it is recommended that
6 be chosen to lie between the values of .05 and .15; thus, the laboratory
personnel are willing to accept out of control data somewhere between 5
and 15% of the time. The economic considerations used for choosing a are
also applicable to the choice of f3.
It is also essential to set maximum and minimum allowable variability
levels. It is necessary to specify a value for the minimum and maximum
amount of variation that will be allowable in the system. These minimum
2 2
and maximum amounts are referred to as o and o respectively. Where
2 ,° }
o =(o-Axo) and
o
o = (a + A x a)2.
1
The values used for Delta (A) should be based on a knowledge of the
variation in the procedure under consideration. However, if such
knowledge is not available A may be arbitrarily set equal to .20.
-------
Mathematical Equations
n 2
I di -
n 2
(I di)
_i
N
N - 1
= Variance of the differences
d
2
S =
o
2
S =
V= Standard deviation of the differences
(.88,) = .64 S, estimates a
d d o
(1.2S.) = 1.44 S, estimates a
d d
(D
UL(M) =
2 log
1
log
.[1 - Bl
^ a J , „
S
. S
1 ' M 1
2
0
1
(2)
2 log.
LL(M)
•f-M
•LI - aJ
log.
s J
o
(3)
-------
Where: UL(M) = upper limit at M sets of samples
LL(M) = lower limit at M sets of samples
di = the difference between the i set of duplicates or
spiked samples
N = the total number of sets of duplicates or spiked
samples used to construct the control charts
2
S = minimum amount of variation allowed in the system
2
Si = maximum amount of variation allowed in the system
a = percent (decimal fraction) of time you are willing
to judge the procedure out of control when it is
in control
6 = percent (decimal fraction) of time you are willing
to judge the procedure in control when it is out
of control
M = number of sets of duplicates or spiked samples used
in calculating the value to be plotted on the chart
-------
CM
•D
Ul
LABORATORY IDENTITY
PARAMETER - METHOD
DATE
RANGE OF CONCENTRATION
oc 8 x? LEVELS
STANDARD DEVIATION
UPPER CONTROL LIMIT EQUATION
LOWER CONTROL LIMIT EQUATION
CONTROL CHART
I I
I I
I L
SAMPLE SET NO. (M)
EFFECT OF a a X? LEVELS ON
STANDARD CONTROL CHART
-------
CONTROL CHART
CALCULATIONS
-------
EXAMPLE I
Accuracy Control Chart
Data:
Actual
.34
.49
.49
.68
.67
.66
.83
.34
.50
.40
.50
.66
.50
.52
.98
.49
1.6
1.3
ioratory A
•zed: Total phosphate
letric with persulfate
12, 1968
phosphorus
digestion
Results of Analyses of Standards
(mg/1 Total PO
Obtained
.33
.49
.49
.65
.65
.70
.80
.34
.47
.40
.53
.60
.56
.59
.75
.63
1.7
1.2
•,-P)
Difference (di)
+ .01
.00
.00
+ .03
+ .02
-.04
+ .03
.00
+ .03
.00
-.03
+ .06
-.06
-.07
+.23
-.14
-.10
+ .10
2
di
.0001
.0000
.0000
.0009
.0004
.0016
.0009
.0000
.0009
.0000
.0009
.0036
.0036
.0049
.0529
.0196
.0100
.0100
-------
Actua^L
3.3
4.9
2.3
1.3
2.3
Obtained
3.3
4.6
2.3
1.3
2.4
Zdi
2
Zdi
2
(Zdi)
.27
.21
.07
Difference (di) di
.00 .0000
+.30 .0900
.00 .0000
.00 _ .0000
-.10 .0100
Calculations
2
Sd =
N - 1
-009
'f
22
-09
.009
(D
2 2
So • ('8V
.64 S, = .64(.009)
a
.006
Si = (1.2SH)^ = 1.44 S^ = 1.44 (.009) - .013
2 loe
UL(M)
•H^-J
*• a J
+ M
-------
CM
T3
Ul
.28
.24
.20
.16
.12
.08
.04
.00
-.04
-.08
-.12
LABORATORY A ACCURACY CONTROL CHART
TOTAL PHOSPHATE PHOSPHORUS - COLORIMETRIC METHOD
WITH PERSULFATE DIGESTION
NOV. 12, 1968
RANGE =.32 to 4.9 mg/l P04'P
a = .15 B - .15
Sd = * .09 mg/l P04 - P
UL =.04 + .008 (M)
LL =-.04 + .008 (M)
I
I
I I
I I
I I I I
3456789 10
STANDARD SAMPLE SET NO. (M)
II
12
13 14
EXAMPLE I - ACCURACY CONTROL CHART
-------
EXAMPLE II
Precision Control Chart
Laboratory: Laboratory AC
Parameter Analyzed: Hexane extractables
Method: Semiwet extraction method
Date: January 5, 1969
Data:
Results of Analyses of Duplicate Samples
Duplicate No. 1
.40
.80
.63
.93
1.46
1.20
1.80
2.16
.40
.20
.40
.46
.40
1.76
.83
1.16
.56
1.26
(mg/1 Hexane
Duplicate No
.50
.83
.60
.83
1.16
1.10
1.56
2.20
.36
.28
.30
.40
.60
1.80
.86
1.02
.63
1.33
Extractables)
. 2 Difference (di)
+.10
+.03
-.03
-.10
-.30
-.10
-.24
+.04
-.04
+.08
-.10
-.06
+.20
+.04
+.03
-.14
+.07
+.07
2
di
.0100
.0009
.0009
.0100
.0900
.0100
.0576
.0016
.0016
.0064
.0100
.0036
.0400
.0016
.0009
.0196
.0049
.0049
-------
Duplicate No. 1 Duplicate No. 2 Difference (dij
.48 .36 -.12
.59 .59 .00
.59 .60 +.01
1.17 1.26 +.09
Zdi = -.470
2
Zdi = .297
2
(Zdi) = .221
Calculations
,H2 (Zdi) .221
-* ~ ~ N " 22 on-
Jd N- 1 21 -°137
2
di
.0144
.00
.0001
.0081
(D
222
S = (.85.) = .64 S, = .64(.0137) = .00877
o d d
222
S = (1.25,) = 1.44 S, = 1.44(.0137) = .01973
id d
UL(M)
1
2
S
o
1
2
S
1
W 1
2
S
o
1
2
S
1
-------
3.5
+ M
.0088
.0088 .0197
_1 _1
.0088 " .0197
3.47 +M .811
63.35 63.35
.054 + .0128(M)
(21
LL(M)
l - a
1
S
0
-3.47
+ M
1
s
1
s
4- M
63.35 63.35
- -.054 + .0128(M)
Upper limits on the Y-axis:
at M = 0
UL (0) = .05 + 0(.013) = .05;
at M = 14
UL (14) = .05 + 14(.013) = .23
Lower limits on the Y-axis:
at M = 0
LL (0) = -.05 -I- 0(.013) = -.05;
at M = 14
LL (14) = -.05 + 14(.013) = .13
(3)
-------
.35
.30
.25
.20
.15
.10
.05
.00
-.05
LABORATORY AC PRECISION CONTROL CHART
HEXANE EXTRACTABLES - SEMI-WET EXTRACTION METHOD
JAN. 5, 1968
RANGE = .20 to 1.8 mg/l HEXANE EXTRACTABLES
oc =.15 /3 =.15
Sd = *.II7 mg/l HEXANE EXTRACTABLES
UL =.054 4 .013 (M)
LL =-.054 * .013 (M)
-.10
-.15
I
I
I
I
I
I
3456789 10
DUPLICATE SAMPLE SET NO. (M)
12 13
14
EXAMPLE 2 - PRECISION CONTROL CHART
-------
USE OF CONTROL CHARTS
Once the control charts are constructed, and prior to their use,
consideration must be given to the number of duplicate analyses to
be conducted during a series of samples; likewise, the same decision
must be made on spiked or standard samples.
In considering the number of duplicate and spiked sample analyses
to be conducted in a series of samples, it is necessary to weigh the
consequences when the data go out of control. The consequences of
this situation are reanalyzing a series of samples or discarding the
questionable data obtained. The samples to be reanalyzed are those
lying between the last in-control point and the present out-of-control
point. A realistic frequency for running duplicate and spiked samples
would be every fifth sample; however, economic consideration and
experience may require more or less frequent duplicate and spiked
sample analyses.
Once the frequency of duplicate and spiked samples has been
determined, it is then necessary to prepare spiked or standard samples
in concentrations relative to the concentration of the control charts,
which should be similar to those of the environmental samples. These
spiked or standard samples must be intermittently dispersed among the
series of samples to be analyzed and without the analyst's knowledge of
concentration. Similarly, duplicate samples must be intermittently
dispersed throughout the series of samples to be analyzed, and ideally,
without the analyst's knowledge; however, this is sometimes very difficult
-------
to accomplish.
The results of the duplicate and spiked sample analyses should
be calculated immediately upon analyzing the samples to allow for
early detection of problems that may exist in the laboratory. An
example of these calculations follows:
Duplicate
Sample No.
M
1
2
3
Results
No. 1 No. 2
5.4
4.8
6.1
5.2
4.7
5.8
2
Difference (di) di
.2
.1
.3
.04
.01
.09
Kdi )
.04
.05
.14
2
Upon plotting the summation or I(di ), one of three possibilities can
occur:
1. Out of control on the upper limit
2. In control within the upper and lower limit lines
3. Out of control on the lower limit
Out of Control on Upper Limit
When data goes out of control on the upper limit the following
steps should be taken:
1. Stop work immediately
2. Determine problems
a. Precision control chart
(1) The analyst
-------
(2) Nature of the sample
(3) Glassware contamination
b. Accuracy control chart
(1) The analyst
(2) Glassware contamination
(3) Contaminated reagents
(A) Instrument problems
(5) Sample interference with the spiked material
3. Rerun samples represented by that sample set number,
including additional duplicate and spiked samples.
4. Begin plotting at sample No. 1 on chart.
In Control
When data continuously fall in between the upper and lower
control limits, the analyses should be continued until an out-of-
control trend is detected.
Out of Control on Lower Limit
When data fall out of control on the lower limit, the following
steps should be taken:
1. Continue analyses unless trend changes
2. Construct new control charts on recent data
3. Check analyst's reporting of data.
-------
ILLUSTRATIONS OF
CONTROL CHARTS
-------
Ul
SAMPLE SET NO.
ANALYSIS IN CONTROL
NO PROBLEMS:
CONTINUE ANALYSIS
w
SAMPLE SET NO.
ANALYSIS OUT OF CONTROL
UPPER LIMIT
PROCEDURES:
I. STOP ANALYSIS
2 LOCATE PROBLEM
3 CORRECT PROBLEM
4 RERUN SAMPLES
5. START CHART AT SAMPLE
SET NO. I.
SAMPLE SET NO
ANALYSIS OUT OF CONTROL
LOWER LIMIT
INCREASED EFFICIENCY OR
FALSE REPORTING
PROCEDURES.
I CONTINUE ANALYSIS
2 CONSTRUCT NEW CHART
WITH RECENT DATA
3. OBSERVE ANALYST
W
SAMPLE SET NO.
ANALYSIS OUT OF CONTROL
UPPER LIMIT
CONTINUOUS ERROR TREND
PROCEDURES:
SAME AS ABOVE BUT STOP
ANALYSIS WHEN TREND IS
DETECTED
LABORATORY
CONTROL
QUALITY
CHARTS
-------
b. Analyze spiked or standard samples intermittently dispersed
among day's samples without analyst's knowledge of concen-
tration.
2
c. Calculate Z(di ) of results as soon as possible
2
d. Plot I(di ) -
(1) Out of control on upper limit -
(a) Stop work
(b) Determine problems
(c) Rerun samples represented by that number
(d) Begin plotting at sample No. 1
(2) In control - continue analyses
(3) Out of control on lower limit -
(a) Continue analyses unless trend changes
(b) Construct new chart on recent data
(c) Check analyst reporting data
e. Compare standard deviation
(1) Other laboratories
(2) Literature
-------
OUTLINE OF PROCEDURES
FOR
CONSTRUCTING AND USING CONTROL CHARTS IN THE LABORATORY
1. Obtain initial sets of duplicate and spiked or standard sample
data for a given parameter under the following conditions -
a. Normal laboratory operations
b. Constant analyst or group of analysts
c. Consistent method
d. Parameter present in a narrow range of concentration
2. Calculate the following -
a. Standard deviation
b. Upper control limits
c. Lower control limits
3. Construct control charts for precision and accuracy. These
charts represent and are restricted to the specific -
a. Laboratory
b. Parameter
c. Range of concentration
d. Analytical method
e. Time
4. Use of control charts -
a. Analyze duplicate samples intermittently throughout
day's samples
-------
The primary reasons for data falling out of control on the lower
limit are increased efficiency or false data reporting.
Standard Deviation
The purpose of calculating the standard deviation is to allow
for inter-laboratory comparisons of precision and accuracy as well
as similar comparisons with the literature. In this respect the
standard deviation can be used as a guide to determine if the
laboratory is operating "in the ball park" on precision and accuracy
for a given parameter. It should be emphasized that the comparisons
of standard deviations should only be used as a guide since the
standard deviation of a specific laboratory is characteristic of that
laboratory's operations and no other.
-------
.0016
.0014
.0012
.0010
.0008
LABORATORY D PRECISION CONTROL CHART
NITRITE NITROGEN - COLORIMETRIC DIAZOTIZATION
.OOO
-.0002k
-.0004
X (.00188)
METHOD
OCT. 10, 1968
RANGE = .020 to .250 mg/l NOg " N
K = .15 x? =.15
Sd = ±.008 mg/l N02'N
UL =.000272 + .000064 (M)
LL =-.000272 + .000064 (M)
<
!
_L
I
I
I
3456789 10
DUPLICATE SAMPLE SET NO. (M)
II
12 13
14
CORRECTIVE PROCEDURES:
I. STOP ANALYSIS AT SAMPLE SET II
2. LOCATE CAUSE OR ASSUME CHANCE CAUSES
3 CORRECT PROBLEM
4 RERUN SAMPLES BETWEEN SAMPLE SETS 10 AND II
5 BEGIN1 PLOTTING I(dE) AT SAMPLE SET I
OU
ONTROL ON UPPER LIMIT
-------
.90
.78
.66
.54
.42
— .30
(V
t3
Ul
LABORATORY B PRECISION CONTROL CHART
B.O.D. ANALYSIS - WINKLER METHOD
OCT. 10, 1968
RANGE = 1.0 to 9.5 mg/l D.O.
<* = .I5 /3 =.15
Sd =4.I8 mg/l D.O.
UL =.13 + .03 (M
LL =-.13 +
.18
.06
-.06
-.18
-.30
3456789 10
DUPLICATE SAMPLE SET NO. (M)
12 13
CORRECTIVE PROCEDURES:
I. STOP ANALYSIS AT SAMPLE SET 13
2. LOCATE CAUSE OR ASSUME CHANCE CAUSES
3. CORRECT PROBLEM IF POSSIBLE
4. SAMPLES CANNOT BE RERUN ON B.O.D. - REJECT
ALL DATA BETWEEN SAMPLE SETS 12 8 13
5. IF DUPLICATES ARE RUN ON ALL SAMPLES -
REJECT SAMPLE SET 13
6. BEGIN PLOTTING Z(d2) AT SAMPLE SET I
14
OUT OF CONTROL ON UPPER LIMIT
-------
Ul
70
60
50
40
30
20
10
0
•10
•20
•30
LABORATORY C PRECISION CONTROL CHART
CHLORIDES - VOLUMETRIC. MERCURIC NITRATE METHOD
SEPT. 4, 1968
RANGE = 10 to 100 mg/l Cl
ex = .15 X? = .15
Sd =*l.5 mg/l Cl
UL = 8.4 + 1.9 (M)
LL =-8.4 + 1.9 (M)
I
I
I I
I I
I 2 3 4 5 6 7 8 9 10 II 12 13 14
DUPLICATE SAMPLE SET NO. (M)
CORRECTIVE PROCEDURES:
I. ANALYSIS COULD HAVE BEEN STOPPED AT SAMPLE
SET 9 OR 10
2. ANALYSIS DEFINITLY STOPPED AT SAMPLE SET II
3. LOCATE CAUSE
4. CORRECT PROBLEM
5. IF ANALYSIS STOPPED AT SAMPLE SET 9 - RERUN
SAMPLES BETWEEN SAMPLE SETS 889
6. IF ANALYSIS STOPPED AT SAMPLE SET 10 - RERUN
SAMPLES BETWEEN SAMPLE SETS 9 S 10
7. IF ANALYSIS STOPPED AT SAMPLE SET II - RERUN
SAMPLES BETWEEN SAMPLE SETS 10 8 II
8. IN ANY CASE BEGIN PLOTTING I(d2) AT SAMPLE SET I
OUT OF CONTROL ON UPPER LIMIT
(CONTINUOUS ERROR TREND)
-------
.28
.24
.20
.16
.12
.08
.04
.00
-.04
-.08
-.12
LABORATORY A ACCURACY CONTROL CHART
TOTAL PHOSPHATE PHOSPHORUS - COLORIMETRIC METHOD
WITH PERSULFATE DIGESTION
NOV. 12, 1968
RANGE = .32 to 4.9 mg/l P04 ' P
a =.15 xf =.15
Sd =±.09 mg/l P04 - P
UL =.04 + .008 (M)
LL =-.04 + .008 (M)
I
I
I
I
I
3456789 10
STANDARD SAMPLE SET NO. (M)
12 13
14
CORRECTIVE PROCEDURES
!. NO PROBLEMS
2. CONTINUE ANALYSIS
IN CONTROL
-------
14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0
2.0
4.0
6.0
LABORATORY A PRECISION CONTROL CHART
AMMONIA NITROGEN - DISTILLATION METHOD
OCT. 10, 1968
RANGE = 1.5 to 6.5 mg/l NH3 - N
a =.15 /
-------
.40
.35
.30
.25
.20
.15
c\T*
ui .10
.05
.00
-.05
-.10
LABORATORY D ACCURACY CONTROL CHART
AMMONIA NITROGEN - DISTILLATION METHOD
OCT. 10, 1968
RANGE = .25 to 6.0 mg/l NH3~N
a = .15 x? = .15
Sd =±.10 mg/l NH3 ~N
UL = .05 + .01 (M)
LL = -.05 •* .0! (M)
I I
I I
I I L I 1 I
I 2 3 4 5 6 7 8 9 10 II 12 13 14
STANDARD SAMPLE SET NO. (M)
CORRECTIVE PROCEDURES:
I. STOP CHARTING AT SAMPLE SET 8
2. BEGIN NEW CHART BY PLOTTING E(d2) OF
SAMPLE SET 8 AT SAMPLE SET I
3. OBSERVE OPERATIONS FOR POSSIBLE PROBLEMS
4. CONTINUE ANALYSIS WITH CAUTION
OUT OF CONTROL ON LOWER LIMIT
(CHANGE OF TREND)
-------
CO
Ul
2.1
1.8
1.5
1.2
0.9
0.6
0.3
0.0
-0.3
-0.6
-0.9
LABORATORY C PRECISION CONTROL CHART
ORGANIC NITROGEN -' KJELDAHL METHOD
NOV. 12, 1968
RANGE = .40 to 2.2 mg/l Organic N
er =.15 /5 =.15
S0 =4.263 mg/l Organic N
UL = .277 + .065 (M)
LL=~.278 + .065 (M)
?T® G
0 © J
0
A A A A
® NOVEMBER
A DECEMBER
I
I
I
I
I
I
234 56 7 89 10 II
DUPLICATE SAMPLE SET NO. (M)
12 13 14
CORRECTIVE PROCEDURES:
I. NOVEMBER DATA IN CONTROL
2. DECEMBER DATA PLOTTED ON SAME CHART
3 CONTINUE ANALYSIS BEYOND SAMPLE SET 6
THROUGH DECEMBER SAMPLES
4. CONSTRUCT NEW CHART ON RECENT DATA
5. PLOT I(d2) ON NEW CHART FOR JANUARY
SAMPLES
OUT OF CONTROL ON LOWER LIMIT
( INCREASED EFFICIENCY )
-------
.080
.070
.060
.050
.040
Ul
.020
.010
.000
.010
.020
LABORATORY D ACCURACY CONTROL CHART
NITRATE NITROGEN - COLORIMETRIC BRUCINE SULFATE METHOD
OCT. 29, 1968
RANGE = .04 to .76 mg/l NOj ~N
« = .15 X? =.15
Sd =±.047 mg/l N03 - N
UL =.0088 + .0020 (M)
LL =-.0088 + .0020 (M)
TRAINING NEW CHEMIST
I I
I
2 3 4 5 6 7 8 9 10 II 12 13 14
STANDARD SAMPLE SET NO.(M)
CORRECTIVE PROCEDURES:
I. SAMPLES I THROUGH 7 ANALYZED BY EXPERIENCED
CHEMIST
2. TRAINING OF INEXPERIENCED CHEMIST BEGAN AT
SAMPLE SET 8
3. TRAINING CONTINUED THROUGH SAMPLE SET II
4. INEXPERIENCED CHEMIST TOOK COMPLETE CONTROL
ON SAMPLE SET 12
EFFECT OF TRAINING ON CONTROL
CHARTS
-------
OPERATING
CONTROL CHARTS
-------
DATA CARD AND MASTER LOG SYSTEM
An analytical laboratory must have an orderly and efficient
system of handling data. This insures the legal defensibility
and validity of the data produced in the laboratory. The FWPCA
has successfully used such a system over the past several years.
Referred to as the data card and master log system, it is composed
of two parts:
1. The data cards for recording all raw data and
computations made by the analyst
2. The master log for recording a summary of
validated data.
The data cards have a consecutive serial number for each
parameter being analyzed. All cards are issued by the laboratory
supervisor and are accountable. The entire operation of arriving
at a value through the various methods of analyses and mathematical
calculations is recorded directly on the data cards, step-by-step.
The analyst is not to recopy raw data from any other source onto the
cards. To insure permanency of these raw data, permanent ink should
be used on the data cards. Completed data cards are to be returned
to the laboratory supervisor for data validation.
The master log is a bound book with pages arranged in original
and tear-out copy order. Page sets are numbered consecutively. The
laboratory supervisor records the validated data in the master log
-------
book. Upon completion of a page in the data book, the tear-out
copy page is removed and used as a working data sheet by the project
director.
Upon completion of a project the numbered data cards and master
log book are stored together for safe keeping and future referral.
-------
ILLUSTRATIONS
of
DATA CARDS
-------
SAMPLE SOURCE
m
DETERMINATION
METHOD
(4)
00000
(2)
ANALYST.
DATA VALIDATED BY: (3)
REFERENCE
DATE
ANAL.
(7)
SAMPLE
NUMBER
(8)
ALIQUOT
(9)
O.D. or
%TRANS.
(10)
Mg- /ALIQUOT
(11)
(12)
FACTOR
(13)
-
Mg/1
(14)
yg/i
(15)
KRC-27 RECORD OF COLORIMETRIC DATA
Key to Annotated Items:
Place in this space
(1) The name of the project or the precise location where sample was collected
(2) The signature of the person analyzing the sample
(3' The signature of person validating data
(4) The parameter being analyzed
(5) The name of the analytical method being used to analyze the sample
(6) The name of the publication that lists the method being used in the
analysis of the sample, such as Standard Methods, 12th edition, 1965
(7) The date the analysis was performed
(8) The number assigned to the individual sample
(9) The number ml. used in the analysis
(10) The optical density or the % transmittance of the sample on a
spectrophotometer
(11) The value of the reading from item (10) taken from a standard curve
(12) Blank space to be used at the analyst's discretion
(13) A number arrived at by dividing the number of ml. of sample used in
the analysis, into the total number of ml. of liquid required for
the analysis, "the Dilution factor"
(14) The value obtained by multiplying item (11) times item (13)
(15) The final value in micrograms/liter if desired
-------
SAMPLE SOURCE R|<$- r^'f^ir/U CC. DETERMINATION /tA?3 - A/
— £7 fcLj_ _£*>.£/• METHOD McD If/ED £i\UC./ME
ANALYST CU^i^ A^ ^-; REFERENCE E P t4 _£>££••/£/ ,41. MtlHDft
DATA VALIDATED BY: r/x-' >><>•/' r^^..^-*^.—
DATE
•?-/* -^J
t,
,,
SAMPLE
NUMBER
»y- /
*,~-3
£~- ^
ALIQUOT
X £ xrv /
///: /
v«<,-,- /
O.D, or
%TRANS.
'£-c C
SL/tf
. /£ C
Mg . /ALIQUOT
i. ^2 /
^, *-J-&
/, 3e
FACTOR
1
/£>
3,
Mg/1
2.3
3^, C
3*6;
ug/1
-
KRC-27
RECORD OF COLORIMETRIC DATA
-------
SAMPLE SOURCE
(11
ANALYST (2}
00000
DATA VALIDATED BY: ^51
DATE
f4)
SAMPLE
NUMBER
(5)
(6)
-
KRC-25
October, 1968
RECORD OF MISC. SAMPLE DATA
Key to Annotated Items Above:
Place in this space
(1) The name of the project or the particular location where sample
was collected
(2) The signature of the person analyzing the sample
(3) The signature of the person validating data
(A) The date the sample was analyzed
(5) The number assigned to the individual sample
(6) The parameter being analyzed
-------
SAMPLE SOURCE
ANALYST
00000
DATA VALIDATtD BY:
ff.
DATE
SAMPLE
NUMBER
s- /
-75
r7,
-jf
(£>.
KRC-25
October, 1968
RECORD OF MISC. SAMPLE DATA
-------
Sample Source
Analvst
C2)
Calculation
'Vtprmin,?r ic n (4)
Me t h C.Q (_s_)_
Rtfertnce (5)
Data Validated BY: r?i
00000
Date
(8)
Sample
Number
(9)
Aliauot
(10)
Flask
Number
(ID
liter
(12)
Blank
Correction
(13)
Corrected
liter
(14)
Factor
(15)
-
Mg./L
(16)
PE1EPM1NATION OF VOLUMETRIC TiTRAfiON
KRC-29
Sept. 1967
Key to Annotated Items Above:
Place in this space
(1) The name of the project or the particular location where sample
was collected
(2) The signature of the person analyzing the sample
(3) A brief formula of the method used to obtain the final mg/1 in
item 16
(4) The parameter being analyzed
(5) The name of the analytical method used to analyze the sample
(6) The name of the publication that lists the analytical method being
used
(7) The name of the person validating the data
(8) The date the sample was analyzed
(9) The number assigned to the individual sample
(10) The number of ml. used in the analysis
(11) The number of the flask or etc. used to titrate the sample
(12) The number of ml. of titrant used in titrating the sample
(13) The number of ml. used to obtain an end point of a blank
-------
(14) The value obtained by subtracting item (13) from item (12)
(15) A number obtained by dividing the number of ml. of sample
used in the analysis into the total number of ml. of liquid
used in the analysis, "the dilution factor"
(16) The value obtained by multiplying item (14) by item (15)
'J.-;A.-g.s CRE
00000
Determiner ii n CML O R \ DE
Method MF.RCUK 1C. titTR/\T,
Analyst
Calculation'^ ; - - (T,-^ - guo A '- X ^
Reference J= flrf ChFlCAL l»; z r i-l c. r>
Data Validated By: __„ /71 _ ^ ,,.- Jt_^-^
Date
r>*>i.
. 5,n/
Flask
Number
1
6
/3
liter
5, JO
?•<"?
6. IO
Blank
Correction
. JO
.JO
.fo
Corrected
Titer
•5'. /O O
<1,05
t>-Oo
Factor
5
/O
JoOO
Mg./L
X?3
9/
bOOG
DETERMINATION OF VOLWETRIC TITRAT10N
KRC-29
Sept. 1967
-------
KRC--26
June 1967
Sample Source (1)
00000
SOLIDS DETERMINATION
Date
(3)
Analyst (4)
DqtP Vfllidaf
SAMPLE
VOLUME
ml
DISH
NUMBER
GROSS
WEIGHT '
gm
ASHED
WEIGHT
gm
TARE
WEIGHT
gm
RESIDUE
gm
VOLATILE
RESIDUE
gm
FACTOR
TOTAL
S. SOLID
Mg/1
T. SUSPENDED
VOLATILE
SOLIDS Mg/1
TOTAL
SOLID
Mg/1
T. VOLATILE
SOLIDS Mg/1
TOTAL D.
SOLID Mg/1
&d Rv: f?1
(5)
(6)
(7)
(8)
(9)
(10)
(ID
(12)
(13)
(14)
(15)
(16)
(17)
(18)
-
-------
Key to Annotated Items on the Solids Card:
Place in this space
(1) The name of the project or the particular location where
sample was collected
(2) The nai^e of the person validating the data
(3) The date the sample was analyzed
(4) The signature of person analyzing the sample
(5) The number assigned to the individual sample
(6) The number of ml. used in the analysis
(7) The number of the container used for the analysis
(8) The weight of the container plus the residue remaining after
drying treatment
(9) The weight of the container plus the residue remaining after
the 600°C heat treatment
(10) The original dry weight of the container
(11) The value obtained by subtracting item (10) from item (8)
(12) The value obtained by subtracting item (9) from item (8)
(13) The value obtained by the formula 777- x 1000
Item (6)
(14) The value obtained by multiplying item (13) times item (ll)
if total suspended solids are analyzed
(15) The value obtained by multiplying item (13) times item (12)
if total suspended volatile solids are analyzed
(16) The value obtained by multiplying item (13) times item (11)
if total solids are analyzed
(17) The value obtained by multiplying item (13) times item (12)
if total volatile solids are analyzed
(18) The value obtained by subtracting item (14) from item (16)
-------
KRC-26
June 1967
Sample Source
/'Aeif.e, T
00000
SOLIDS DETERMINATION
/vli£ /? Date 3 ~ 6? ~
Analyst
Validated By:
SAMPLE
VOLUME
ml
DISH
NUMBER
GROSS
WEIGHT
grc
ASHED
WEIGHT
gm
TARE
WEIGHT
gm
RESIDUE
gm
VOLATILE
RESIDUE
gm
FACTOR
TOTAL
S. SOLID
Mg/1
T. SUSPENDED
VOLATILE
SOLIDS Mg/1
TOTAL
SOLID
Mg/1
T. VOLATILE
SOLIDS Mg/1
TOTAL D.
SOLID Mg/1
£-/
/c
$
tf-fo/O
33.^'C'
3?.3crC\ i'-C (.'
/t:,/l'C
/*•'', L'L C'
C "
b "'-
^25
/7
.23.fa/L>
&• y^te
31..S//C
, y/ r r
. £> OOO
^-C'.OO'P
/t. L/C c
0
S-3
/c c
3
2£< /7$3
& /we
33.,'ff
,c?SS
. c?% 3
/
-------
Rev.
KRC-28
Mar. 1968 BIOCHEMICAL OXYGEN DEMAND AT 20'C
ooooo
SAMPLE SOURCEL
(1)
DATA VALIDATED BY:
DATE_
(3)
AHALYST__IA1.
(2)
SAMPLE
Z CONCEN-
TRATION
DAYS
INCUBATED
TIME
BOTTLE t
DISSOLVED
OXYGEN
INITIAL
BOTTLE #
DISSOLVED
HYV^FM
FINAL
ACTUAL
DEPLETION
BLANK
CORRECTION
CORRECTED
DEPLETION
DILUTION
FACTOR
B.O.D.
mg/1
B.O.D.
mg/1
(5)
(6}
(7)
(8)
(9)
(10)
(11>^
(J2-)
(13)
(1A)
(15)
(16)
(17)
(18)
(19)
/
/
/
\/
,
/
[>
/
/
*
/
-------
Key to Annotated Items on Biochemical Oxygen Demand Card:
Place in this space
(1) The name of the project or the particular location the
sample was collected
(2) The nane of the person validating the data
(3) The date the BOD was set up
(4) The signature of the person analyzing the sample
(5) The number assigned to the individual sample
(6) The % dilution of the sample
(7) The number of days the sample incubated
(8) The time the BOD was set up
(9) The bottle number of the initial DO
(10) The value of the initial DO in mg/1
(11) The bottle numbers of the two samples to be incubated
(12) The final DO value of the two incubated bottles
(13) The average of the values in item (12)
(14) The value obtained by subtracting the value of item (13)
from item (10)
(15) The seed correction value. Used only when the sample
was seeded
(16) The value obtained by subtracting item (15) from item
(17) The value obtained by dividing the % sample used into 100%
(18) The value obtained by multiplying item (17) by item (18)
(19) The BOD value to be reported
-------
00000
KRC-28
July 12, 1968 BIOCHEMICAL OXYGEN DEMAND AT 20°C
Sample Source ^J/i M pg Q CVgpX. Date g - / O - r
Zv>
S.fS
P^
^
/.ci
1./3
.-=_
7/3
/
7/.3
7
5- /
£~£>
•5"
',— ->
^6
f.jo
*y£
/^v1'
&*
4,£c
1..50
—
^ £T<"
2
7' 0 0
$-1
£5
$
34 A
%.J>0
^
^
6.4S'
/- 75
—
/, 7. 6"
4-
?< ? 0
S-2
SO
5
4-/0
7- 95
^
<$/
^
_____
-^
. —
S-2
^5"
5
•5/4T
7-9O
^
^
3>bo
.-T.-Pr
- —
5:5^
5
//#
g-70
ic^
^5fr
^
/<*'^
£.5 (^
3-3 G
• —
-3^-^
S&
M-€-
-------
Card #1
SAMPLE SOURCE_
ANALYST
CD
(2)
DETERMINATION
METHOD (4)
Metals
Data Validated by: (3)
Concentrations expressed in yg/1 or mg/1 (5)
Date
Anal.
(6)
Samp .
No.
(7)
Fact.
(8)
As
(9)
Be
Ca
Co
K
Li
Mp
NA
Se
Tl
V
-
Sb
Card
SAMPLE SOURCE (1)
ANALYST (IT
DETERMINATION_
METHOD
Metals
Data Validated by:
Concentrations expressed in pg/1 or mg/1
Date
Anal.
(6)
Samp.
No.
(7)
Fact.
(8)
Zn
(9)
Cd
B
Fe
Mo
Sn
Mn
Cu
Ap
Ni
Al
Pb
Cr
Ba
Sr
RECORD OF METALS ANALYSES
Card # 2 can be used alone or as a continuation for Card /•' 1
-------
Key to Annotated Items on Record of Metals Analyses:
Place in this space
(1) The name of the project or the precise location where sample
was collected
(2) The signature/s of the person/s analyzing the sample/s
(3) The signature of the person validating data
(4) The method used to analyze the sample/s
(5) Indicate the unit in which concentrations are expressed
(6) Date of analyses
(7) The sample number or code
(8) The sample dilution or concentration record
(9) The element analyzed
-------
Card #1
SAMPLE SOURCE
ANALYST (L^.
K*tV l~ /.'
DETERMINATION
METHOD
Metals
Data Validated by : -
,
<^ - Concentrations expressed irf'^ig/l^or mg/1
Date
Anal.
3-3'7l
Samp .
No.
dkc-~
Fact.
/
As
Be
- /3
Ca
/.V5
/
Co
K
Li
Mp
y.jy
NA
6.'/3*
Se
Tl
V
-
Sb
Card #2
SAMPLE SOURCE ,
ANALYST It
.7"/£ x'"") U g A?
Data Valids'ted by:
DETERMINATION
METHOD /s'SCl
Metals
Concentrations expressed in(^/T;or mg/1
Date
Anal.
S-3-7/
Samp.
No.
X/^r-^
Fact.
/
Zn
Cd
B
Fe
Mo
Sn
Mn
Cu
t-M
Ap
Ni
Al
Ph
Cr
.;•'*
Ba
Sr
RECORD OF METALS ANALYSES
Card r 2 can be used alone or as a continuation for Card
-------
LABORATORY SCHEDULE AND DATA RECORD
The "Laboratory Schedule and Data Record" as illustrated in Figure 1
consists of an original and three copies of NCR paper. It serves a
twofold purpose:
(1) It lists the parameters that the laboratory
is requested to analyze.
(2) It serves as a permanent record for the data,
listing all pertinent information about each
sample/s.
The "Schedule" is completed by the person/s requesting analytical services,
and is delivered to the laboratory with the samples. After the laboratory
personnel validate the data on the data cards, the data are transferred to
the "Schedule" and forwarded to the chief of the laboratory for a final
review. Upon his approval the original is filed in the "Master Data Log".
The first copy is forwarded to the person who requested the analytical
services; the second copy is forwarded to STORET (if applicable); the third
copy is retained by the person requesting the data when he delivered the
samples to the laboratory.
-------
-------
Key to Annotated Items:
Place in this space
(1) The name of the project or the precise location where the sample
was collected
(2) Signature of the person who received the saraple/s in the laboratory
(3) The signature/s of the person/s who collected the sample/s
(4) The signature of the person making the final review of the data
(5) The date of the final data review
(6) The code or number assigned to the sample/s
(7) A precise description of where the sample was collected
(8) The date the sample was collected
(9) The time the sample was collected
(10) The date the sample arrived in the laboratory
(11) The latitude and longitude of the sampling point (if desired)
(12) The name of the parameter to be analyzed
(13) A check mark if it is desired that this sample be analyzed for this
parameter
(14) The value of the analysis for this parameter
-------
K
r •'
-4
r-
Z '•
i
s».
- '
Tl
" !
J '
H
r
n ,
s (
_ i
3-
> (
TI ,
=
D •
|'i
H
/I
_< ,
i
p
5 •
< i
3
£
T| _
") r
D,
3
-1 '
> i
Z ,
~) ' '
? '
S 'i
A
i i|
r
;
i
II
II
1
1
L.
.
|
r
\
K
^,
-
\
•s
N,
^
\
K
(^-
*\
^
Ui
'^
^
s.
L*.
1
\
V^J
\
Ul
\
:S
«-
^
\
A
^
\
^
oc
1
"Z.
[-
;.
"^
^
T-
\
In
^
*•
-
•
',-
r
"
\
I
LONGITUDE
o
r
O
0
,
-I—. 1
^_S-
o
.
o
0
0
,
0
r
O
LATITUDE
0
-
.
o
,
-*
0
.
o
.
o
=
o
0
.
o
.
LAB ARRIVAL DATE
C
C
1
I—1
o
o
o
1
o
-0
>-
0
I
0
t—l
^-
o
1
0
0
o
1
t
-i
£
o
o
o
0
10
.u
t—l
o
o
o
o
D
0
r~
m
n
6
z
o
o
o
1
s
H-
0
0
o
^_,
^_.
o
o
1
(—•
o
o
L
•-J
*-
"
; o
O trt 1 O
w s ^
ri ^ i
X 0 i' 1
— z f c ;
6 ! m
2 ,.-3 j
g r s ]
vj _ o p
c- re
C* H- 93 !
o" ' re i
S f — • <
re o re cs
7T E H- i-i s: ]
0 3 i-
c. M- re
cr 5: to 93 i
it, re ui H' i j
Mi Hi • < C3
H- re 93
« 3 O t-j
i- re i- to
H h- g i
X P- i I
Z. i
re j
* rr 4 NJ 33 i
N) rt) re UH ^«
Ml Ml • < C=
H- H- H- re ^
C 3 » 1 '•&•
S S 0 S i
H- t—
o
CC W 93
»1 O H-
IK re c
i-i l^-.
S^K |-
^ re
K ?•
re
E h
re 1
j
i
1 i
|
;
i
1
a: -t p>
SB5
pi X —
Z 2 5
n i
(Tl t>
5 S
n-. 2
a; n
03 j
O i
x> ,
t
f- i
b i
h- u,
s
•o
O
X
m
o
c
t.
z
o
D
-f
L
loo
m
n
o
lo1
i
-------
ADDENDA SHEET
TO THE PUBLICATION:
"AN ANALYTICAL QUALITY CONTROL PROGRAM FOR EFFICIENT LABORATORY MANAGEMENT"
The following paper, "An Analytical Quality Control Program for
Efficient Laboratory Management," was written for presentation at the
20th Annual Oklahoma Industrial Waste and Pollution Control Conference,
Oklahoma State University, Stillwater, Oklahoma, March 31-April 1, 1969.
It should be noted that several name and title changes have occurred
since the paper was published in 1969:
Federal Water Pollution Control Administration,
U. S. Department of the Interior is now the
U. S. Environmental Protection Agency;
Mr. R. E. Crowe is now attached to the Research and Development
Program, U. S. Environmental Protection Agency, Washington, D.C.
Dr. R. Harkins is now Mathematical Statistician, Ada Facility,
Surveillance and Analysis Division, U. S. Environmental Protection
Agency, Ada, Oklahoma;
Mr. J. Kingery is now Mathematical Statistician, Ada Facility,
Surveillance and Analysis Division, U. S. Environmental Protection
Agency, Ada, Oklahoma, and
Mr. B. G. Benefield is now Chemist, Ada Facility, Surveillance
and Analysis Division, U. S. Environmental Protection Agency,
Ada, Oklahoma.
-------
AN ANALYTICAL QUALITY CONTROL PROGRAM
FOR EFFICIENT LABORATORY MANAGEMENT*
by
R. E. Crowe, R. Harkins, J. Kingery, and B. G. Benefield**
Introduction
Quality control procedures in general have been used since man
began his thinking process. Galileo, in his experiments to determine
the surface tensions of liquids, gave detailed instructions for ob-
taining a consistent set of results (1). The artisan guilds of the
Middle Ages prescribed extended apprenticeships before a person was
considered a master craftsman. This training maintained a level of
competence within the guild (2). Dr. A. Shewhart of Bell Telephone
Laboratories developed the basic theory of control charts in the
1920s (3). This was the beginning of industrial use and acceptance of
these and other statistical techniques to measure the quality of prod-
ucts of a manufacturing process. The development of these techniques
in industry led to their limited use in the analytical laboratory (4).
* A paper scheduled for presentation at the 20th Annual Oklahoma
Industrial Waste and Pollution Control Conference, Oklahoma State
University, Stillwater, March 31-April 1, 1969.
** Respectively, Chief, Chemistry and Biology Section, Technical
Assistance, Technical Services Program; Acting Chief, Pollution
Surveillance, Technical Services Program; Mathematical Statistician,
Pollution Surveillance, Technical Services Program; and Chemist,
Chemistry and Biology Section, Technical Assistance, Technical
Services Program, all of the Robert S. Kerr Water Research Center,
Federal Water Pollution Control Administration, U. S. Department
of the Interior, Ada, Oklahoma.
-------
Although laboratory operations are not considered to be manufactur-
ing processes, it can be recognized that the analytical data produced
by any laboratory are, in actuality, the products of that process. As
•with industry, a quality control program should be employed in the labora-
tory to insure the quality of its products, which in turn, characterize
the normal laboratory operations, and detect abnormal operations when
they occur. This paper presents one such program for consideration as a
tool in characterizing a laboratory's operations and maintaining quality
control in the laboratory.
Laboratory "Fingerprint"
Anyone who has worked in, supervised, or managed the operations of
an analytical laboratory is well aware of the basic tools for determining
quality of data produced. These tools are duplicate sample analyses and
spiked or standard sample analyses. These have been used to indicate the
precision and accuracy of the process producing the data.
Those who have used these tools recognize the difficulty involved in
making a decision as to the validity of the data produced based on dupli-
cate and spiked or standard sample analyses. The common practice has been
to visually observe the data and arbitrarily judge their acceptability
with no concrete basis for the decision. Obviously, it would be advan-
tageous to have a so-called "fingerprint" of the precision and accuracy
for the normal operations of a specific laboratory group for the analysis
of a specific parameter.
-------
"Fingerprints" of this nature can tell us many things about the
laboratory's operations. For example, they could tell us when problems
exist with the analysts, reagents, glassware, instruments, etc. They
could indicate whether the laboratory is operating normally or abnormally,
thus pointing out when data generated should be accepted, questioned, or
rejected. In addition, they could tell us when the laboratory is operating
at optimum efficiency. As with industry, the laboratory sometimes gener-
ates products which are not acceptable. In these cases, samples must be
analyzed again to produce acceptable results. These "fingerprints" could
help us determine which samples or sets of samples should be reanalyzed.
The laboratory "fingerprints" we refer to are in the form of control
charts. The construction and use of such charts are discussed below.
Construction of Control Charts
As we have indicated, two control charts are required to "fingerprint"
the laboratory operations for a given analytical procedure. These are
referred to as precision and accuracy control charts. Precision control
charts are constructed from duplicate sample analyses data; accuracy
control charts are constructed from spiked or standard sample analyses
data. A set of the two represents, and is restricted to, a specific
laboratory, group of analysts, analytical method, range of concentration
and period of time. To construct the precision and accuracy control
charts, it is necessary to obtain several sets of duplicate and spiked
-------
or standard sample data. The greater the number of sets of initial data
obtained, the better the "fingerprint" of the laboratory operations. Eco-
nomics must be considered, however, in obtaining the initial sets of data.
It is recommended that at least 20 sets of duplicate and 20 sets of spiked
sample data from an in-control process be used to initially construct the
control charts. The selection of in-control data can be a judgment decision
or, if desired, extreme values can be systematically eliminated by the
Dixon and Massey method (5) of processing data for extreme values.
The initial sets of data must be obtained under the following condi-
tions :
1. Normal laboratory operations
2. Constant analyst or group of analysts
3. Consistent method
4. Narrow range of concentration of the
parameter analyzed.
The reasons for the first three conditions are obvious; number A needs
more explanation.
The precision and accuracy of the analyses for many parameters are
proportional to the concentration of the parameter to be measured. This
may require the use of several control charts in varying ranges of
concentrations for a given parameter. Duly experience will dictate this.
It is important to note here that there is complete control over the
range of concentration of spiked samples or standard samples, but little
-------
or no control over the range of concentration of the duplicate samples.
It is important also to point out the basic differences between a
spiked and a standard sample. These terms have been used synonymously
at times in the discussion of accuracy data; however, the two differ
greatly. A spiked sample can be defined as an environmental sample to
which has been added or "spiked" a known quantity of that parameter
already present in the sample in significant concentrations. The envi-
ronmental sample obviously must be analyzed before as well as after the
"spiking" of the sample. Spiked samples should be used in situations
where knowledge is insufficient as to the interferences of the method or
of the environment from which the sample was obtained. This situation
might also necessitate a complete and independent study of the inter-
ferences .
A standard sample can be defined as one prepared by adding a known
concentration of a given parameter to distilled water. The sample
should then be analyzed identically to the environmental samples. A
standard sample can be used in situations where the interferences of
the method with the environment are sufficiently known. In other words,
a standard sample can be used where interferences of the method are not
questioned, and the assumption made that interferences are absent. A
standard sample has the inherent disadvantage of unusual appearance, so
that the analyst is aware of its introduction into a series of environ-
mental samples. This could create bias in the results.
-------
The control charts are derived from three basic calculations. No
attempt is made here to develop the mathematics upon which the control
charts are based. However, references are given so that those who wish
to delve more deeply into the subject may do so (6).
These basic calculations are:
1. Standard deviation of the differences between
duplicates, or in the case of spiked or standard
samples, between the known quantity and the
quantity obtained.
2. The upper control limit
3. The lower control limit.
Prior to these calculations, two decisions must be made:
1. The a and 3 levels
2. The allowable variability levels
By definition, a is the probability of judging the process to be
out-of-control, when in fact, it is in-control. It is recommended that
a be chosen to lie within the boundaries of .05 and .15; that is, the
laboratory personnel are willing to stop the laboratory process some-
where between 5 and 15 percent of the time, judging it to be out-of-control,
when in fact, it is in-control. If the cost of examining a process to
determine the reason or reasons for being out-of-control is considerable,
then it may be desirable to choose a low a. Likewise, if the cost is
negligible, it may be desirable to choose a larger a value and thus stop
the process more frequently.
-------
On the other hand, t? is defined as the probability of judging the
process to be in-control when it is not. Again, it is recommended that
g be chosen to lie between the values of .05 and .15; thus, the laboratory
personnel are willing to accept out-of-control data somewhere between 5
and 15 percent of the time. The economic considerations used in choosing
a also apply in choosing 6. The effects of varying a and 6 are demon-
strated in Figure 1.
It is also essential to set maximum and minimum allowable variability
levels. It is necessary to specify a value for the minimum and maximum
amount of variation that will be allowable in the system. These minimum
2 2
and maximum amounts are referred to as o and o respectively. The values
used should be based on a knowledge of the variation in the procedure under
consideration. However, if no such knowledge is available, the values
2 22 2
may be arbitrarily set at o = (a - .20a) and a = (a + .20a) .
0 1
n 2
Z di -
n 2
(£ di)
2 1=1
S, = - = Variance of the differences
N - 1
S, = V^d = Stan^ard deviation of the differences (1)
22 2
S = (.85.) estimates a
o d o
22 2
S = (1.2S,) estimates c
1 d 1
-------
UL(M) =
'[M
LL(M) =
2
l°ge
1
6
LI -
1
2
1 e
J , .,
Si
-F
u o J
' M 1 1
Where: UL(M) = Upper limit at M sets of duplicate or spiked samples.
LL(M) = Lower limit at M sets of duplicate or spiked samples.
di = The difference between the i set of duplicates or
spiked samples.
n = The total number of sets of duplicates or spiked samples
used to construct the control charts.
2
S = Minimum amount of variation allowed in the system.
2
S = Maximum amount of variation allowed in the system.
a = Percent of time you are willing to judge the procedure
qut-of-control when it is in-control.
6 = Percent of time you are willing to judge the procedure
in-control when it is out-of-control.
M = Number of sets of duplicates or spiked samples used in
calculating the value to be plotted on the chart.
For clarification purposes, an example of using the above equations in
making the calculations is given below. The example involves the measure-
ment of total phosphate phosphorus by the colorimetric, with persulfate
digestion, method. Twenty-three sets of standards at concentrations
-------
varying from .32 to 4.9 mg/1 of total phosphate phosphorus were used
in the calculations. It was assumed that there was no appreciable
proportional error in this range of concentration. Also, by visual
observation we did not reject any data as being out of control.
Actual
.34
.49
.49
.68
.67
.66
.83
.34
.50
.40
.50
.66
.50
.52
.98
.49
1.6
1.3
3.3
4.9
2.3
1.3
2.3
Results of Analyses
(mg/1 Total
Obtained
.33
.49
.49
.65
.65
.70
.80
.34
.47
.40
.53
60
-56
.59
.75
.63
1.7
1.2
3.3
4.6
2.3
1.3
2.4
of Standards
P04-P)
Difference (di)
+ .01
.00
.00
+ .03
+ .02
-.04
+ .03
.00
+ .03
.00
-.03
+ .06
-.06
-.07
+.23
-.14
-.10
+ .10
.00
+.30
.00
.00
-.10
Mi
.0001
.0000
.0000
.0009
.0004
.0016
.0009
.0000
.0009
.0000
.0009
.0036
.0036
.0049
.0529
.0196
.0100
.0100
.0000
.0900
.0000
.0000
.0100
-------
Edi
(Zdi)
N
Zdi =
Zdi2 =
(Idi)2 =
.07
.27
.21
.07
,21 -
N - 1
22
.009
V
.009
.09
(1)
2 2
S, - (.8SJ
.645, = .64(.009) = .006
Q
S = (1.2SJ = 1.44S, = 1.44(.009) = .013
a a
2 log
&
a J
UL(M) =
M
o J
3.5
log.
+ M
.013
e L.006
1
.006
1
.013
.006 .013
3.5 . M .69
-r M
90
90
.039 + .0077(M)
(2)
10
-------
LL(M)
2 log -r-
e L 1
1
2
S
0
J? 1
- a J
1
2
S
i
log£
+ M ™»
o •'
PI
1 1
2 2
S S
0 ]
90 90
= -.039 + .0077(M) (3)
We are now prepared to construct an accuracy control chart. The
upper limits on the Y-axis can be calculated using equation (2):
at M = 0
UL (0) - .04 + 0(.008) = .04;
at M = 14
UL (14) - .04 + 14(.008) = .15
These two points can now be plotted to form the upper limit line
as shown in Figure 2.
The lower limits on the Y-axis can be calculated using equation (3):
at M = 0
LL (0) • -.04 + 0(.008) = -.04;
at M = 14
LL (14) = -.04 + 14(.008) = .07
These two points can now be plotted to form the lower limit line
as shown in Figure 2.
It should be noted that the Y-intercept for the lower control line
is the negative of that for the upper control line. This is because
a and 3 are equal. If they are not equal, this condition will not exist.
11
-------
Figure 2 now represents an accuracy control chart for total phosphate
phosphorus which is characteristic of the laboratory operations restricted
to the conditions specified on the chart, Only an accuracy control chart
has been demonstrated here. The same procedures should be followed to
produce a precision control chart.
Use of_ Control Charts
Now that the control charts have been constructed, we are prepared
to plot the values obtained from duplicate and spiked sample results from
a series of sample analyses. At this point a decision must be made as
to the number of duplicate analyses to be conducted during a series of
samples; the same decision must be made on spiked or standard samples.
This decision is primarily one of economics.
In considering the number of duplicate, and spiked sample analyses
to be conducted in a series of samples, it is necessary to weigh the
consequences when the data goes out-of-control. The consequences
involve reanalyzing a series of samples, or discarding the question-
able data obtained. The samples to be reanalyzed should be those lying
between the last in-control point and the present out-of-control point.
For example, if you have a 100 sample series to be analyzed, and a
duplicate and spiked sample are analyzed only once in the series (in the
area of the 50th sample) the consequences, if this one sample is out-of-
control, are that the first 50 samples must be reanalyzed or discarded.
12
-------
If all 100 samples were analyzed prior to calculating and plotting the
out-of-control samples, then all 100 would need to be reanalyzed or
discarded. If duplicate and spiked samples are analyzed more frequently
and more realistically, such as at every fifth sample, then it is
apparent that, if one sample goes out-of-control, it is necessary to
reanalyze only the nine in between the two in-control points, or only
five if the laboratory operations are halted at the out-of-control point.
In addition, the more frequently the duplicate and spiked samples that
are analyzed, the greater the chances of detecting abnormal operations
as they occur. Also to be considered are the economics involved in the
method used and the stability of the sample. A good example lies in the
biochemical oxygen demand (BOD) analysis. It is obvious that this method
requires five days, and the sample could not be reanalyzed after that
time lapse. For these reasons we conduct duplicate analyses on each BOD
analysis and report only those that are in-control.
Once the frequency of duplicate and spiked samples has been deter-
mined, it is necessary to prepare spiked or standard samples in concen-
trations relative to those of the control charts which should be similar
to concentrations of the environmental samples. These spiked or standard
samples must be intermittently dispersed among the samples of the series
to be analyzed and without the analyst's knowledge of concentration.
Similarly, duplicate samples must be intermittently dispersed throughout
13
-------
the series of samples to be analyzed, and ideally without the analyst's
knowledge; however, this is sometimes very difficult to accomplish.
It cannot be overemphasized that the results of the duplicate and
spiked samples must be calculated immediately upon analyzing the
samples. This will allow the Z(d2) to be plotted as soon as possible on
the control charts so that any existing problems can be corrected and
samples promptly reanalyzed. A brief and simplified example of these
calculations is:
Duplicate
Sample No. Results
M No. 1 No. 2
1 5.4
2 4.8
3 6.1
5.2
4.7
5.8
Difference
.2
.1
.3
(d) d2
.04
.01
.09
I(d2
.04
.05
.14
Following each calculation, the summation or £(d2) is plotted on a
chart similar to Figure 2, plotting £d2 against the sample number. Upon
plotting £(d2) one of three possibilities will occur:
1. Out-of-control on the upper limit
2. In-control within the upper and lower
limit lines
3. Out-of-control on the lower limit
Each of these possibilities will now be discussed in detail.
There are generally two types of out-of-control on the upper limit
conditions. One type is illustrated in Figure 3. This occurs when
the laboratory is operating quite normally and suddenly a point goes
14
-------
out-of-control, usually extremely far from the upper control line. The
other type of condition is one in which you have a continuous error trend;
in other words, you are approaching an out-of-control condition at a con-
sistent rate, usually foriring a trend line in the direction of and at an
angle to, the upper control limit line as illustrated in Figure I*. The
latter condition (Figure 4) is advantageous over the former since problems
can usually be detected early in the continuous error trend condition and
corrected before out-of-control actually occurs. The former (Figure 3)
case yields no such warning; the operations go out-of-control with no
prior indication of a problem. When the operations go out-of-control
at the upper limit, obviously the laboratory operations are to be stopped
as soon as possible so that the problems can be located t^-\ corrected
before proceeding with the analysis. Also, keep in mind that it is
possible to go out-of-control for no other reason than chance causes.
Then the samples in question are to be reanalyzed with duplicate and
spiked samples. The first duplicate and spiked sample data are to be
plotted, beginning with Sample Ko. 1 on the control chart. The primary1
reason for starting at position No. 1 on the control chart is that, since
we are plotting the summation of the differences uared, the out-of-
control point dominates the next point, thus continuing out-of-control
even though it could actually be in-control.
What problems are associated with laboratory operations being
15
-------
out-of-control? The answer depends upon whether it is a precision
control or an accuracy control chart. The causes for out-of-control
on the precision chart are usually one or more of the following:
1. The analyst
2. Nature of the sample
3. Glassware contamination
Obviously, we are not concerned over reagents or instrumentation
in precision, since analyzing a duplicate sample would duplicate any
reagent or instrument error. The analyst is probably the primary source
of precision errors; however, the nature of the sample is not to be
overlooked. By the nature of the sanrole we mean the homogeneitv of the
sample in relation to its anenabilitv to being separated into two equal
parts to allow a true duplicate analvsis. In the case of samples that
contain oils or clumps of insoluble material, it is practically impossible
to obtain a duplicate sample. Glassware contamination is probably the least
frequent contributor to imprecision; however, it does occur. One
flask can be contaminated, where another is not.
Six problems are associated, singly or in combination, with an
out-of-control condition on an accuracy control chart. They are:
1. The analyst
2. Glassware contamination
3. Contaminated reagents
4. Instrumentation
15
-------
5. Sample interference with the spiked material
6. Contaminated laboratory atmosphere
All weigh fairly evenly as possible causes of inaccuracy.
The second possibility is where the laboratory is operating in-
control within the upper and lower limit lines. This possibility is
illustrated in Figure 5. Under these conditions there is no cause for
concern over the quality of the data, and samples should be continually
analyzed until either a trend develops or a result goes out-of-control
at the upper limit.
The remaining possibility is out-of-control on the lower limit as
illustrated in Figure 6. This situation is indicative of greater precision
and accuracy being attained by a laboratory. This probably would show
up in the first plotting on the first control chart, because the more
experience the laboratory gains in analyzing a specific parameter by a
specific method, the more precise and accurate that laboratory becomes
until optimum operation is achieved. When this situation occurs, it is
not necessary to stop the analyses; instead, analyzing should continue on
the particular sample series involved (unless the trend changes signigicantly)
It is then necessary to construct a new control chart using the latest
duplicate and spiked sample data. The new control chart will then rep-
resent the current operating characteristics of the laboratory at that time.
In the situation of a new laboratory analyzing a new parameter using an
unfamiliar method, several charts may be constructed before optimum
17
-------
operating conditions are attained.
Another reason for an out-of-control on the lower limit occurrence
would be the analyst's reporting of false data. This is particularly
true with duplicate sample analyses where the analyst is aware of the
duplicate sample. This would not be the case with spiked sample analyses
since the analyst would not have knowledge of the concentration of the
parameter in the standard or spiked sample. If the analyst is suspected
of reporting false duplicate data, it would be necessary to mask the
duplicate samples so that the analyst is not aware of their presence.
It should also be pointed out that analyzing a duplicate or spiked
sample many times with special attention will produce more precise and
accurate data than under normal operations. This, in turn, would produce
an out-of-control on the lower limit condition.
Again, there will be cases where data are out-of-control for no
apparent reason. Many such cases can be attributed to chance causes
which will occur occasionally.
Standard Deviation
Mentioned previously were the three basic calculations required to
"fingerprint" a laboratory's operations. We have discussed the upper
and lower control limits. Now let us briefly discuss the third calcu-
lation, that of the standard deviation. The purpose of calculating the
standard deviation is to allow inter-laboratory comparisons of precision
18
-------
and accuracy. It also allows similar comparisons with the literature.
In this respect, it can be used as a guide to determine if the laboratory
is operating "in the ball park" on precision and accuracy for a given
parameter. It should be emphasized that the comparisons of standard
deviations should be used only as a guide since the standard deviation
of a specific laboratory is characteristic of that laboratory's operations
and no other.
Summary
Some type of analytical quality control program is necessary for
efficient laboratory management and to validate analytical data. Since
analytical data are the products of the analytical laboratory, we must
know which products to reject and which are to be accepted as valid.
The quality of the laboratory products will vary as long as humans are
involved in the analyses; therefore, it is important to know when the
variations go beyond those occurring under normal laboratory operations
so that the end product quality is known.
This paper discusses the construction and use of laboratory "finger-
prints", in the form of control charts, to identify and characterize the
operations of a laboratory. These "fingerprints" or control charts are
limited to the laboratory from which the data are produced. They are
also restricted to the conditions under which the samples are analyzed.
They were constructed intentionally with the use of basic mathematical
equations, thus encouraging their use.
19
-------
The uses of control charts are detailed. Three possible
occurrences when plotting analytical data on the control charts are
described, these being:
1. Out-of-control on the upper limit
2. In-control
3. Out-of-control on the lower limit
The charts are interpreted and recommendations made as to what to do
when the laboratory operations deviate from normal.
20
-------
Kl
LABORATORY IDENTITY CONTROL CHART
PARAMETER METHOD
DATE
RANGE OF CONCENTRATION
<* a 0 LEVELS
STANDARD DEVIATION
UPPER CONTROL LIMIT EQUATION
LOWER CONTROL LIMIT EQUATION
i I I I I 1 I \ 1 I I I I 1
SAMPLE SET NO.(M)
FIGURE i - EFFECT CF VARYING a 8 X?
-------
L-'-E^ATORV A ACCURACY CONTROL CHART
TT;.^ pHv"cPKAT£ PHOSPHORUS
CCOMM."P'C METHOD WITH PERSULFATE
DIC-ESTiON
I'OV. I?, 1968
RiNS-l = .SI tc <.S n;/! P04'P
fi = .!l> x-f s.15
Sj, «».OS m;/| PC^ - P
ULIM}«.04 « .COS (M)
.;.(M',r -.04 * .COB (W!
r 7 6 S
SAf/"LE SET N'O.(M)
_. J . , ,
10 II 12 iS 14 15
-------
.9 0
.78 -
.66 -
-.06 -
-.18-
-.30 -
LABORATORY B PRECISION CONTROL CHART
B.O.D. ANALYSIS WINKLER METHOD
OCT. 10, 1968
RANGE - 1.0 to 9.5 mg/l D.O.
Sd =«.I8 mg/l D.O.
UL(M)».I3 » .03 (M)
LL(M)«-.I3 * .03 (M)
I I I I T I i I I I 1 | [ \
I 2 3 4 56 78 9 10 I I 12 13 14 15
DUPLICATE SAMPLE SET NO.(M)
FIGURE 3 - OUT-OF-CONROL ON UPPER LIMIT
-------
70
60-
50-
40-
30-
20-
- ion
•20-1
-30-
LABORATORY C PRECISION CONTROL CHART
CHLORIDES VOLUMETRIC, MERCURIC NITRATE
SEPT.
RANGE
O-s.15
Sd sil
UL(M)«
LL(M)«
1 I 1 1 1 1 II
12345678
4, 1966
= 10 to
X? =.15
5 mg/l
8.4 t
-8.4 «
1
9
100 mg/l Cl
Cl
1.9 (M)
1.9 (M)
1 I 1 1 1
10 II 12 13 14 I!
DUPLICATE SAMPLE SET NO.(M)
FIGURE 4 - OUT-OF-CONTROL ON UPPER LIMIT
( CONTINUOUS ERROR TREND)
-------
.24 -r
.20i
6 -
".08 -'
-.16
>pv A ACCU-iC-
TOTAL PHOSpHATE PHCS-'",
COLOP.IMETRIC METHOD W;~
DIGESTION
NOV. 12, 1968
RANGE * .32 to 4.9 ng/l !-:
a s.i5 /« =.15
Sd = «.09 mg/l P04 - P
UL(M)«.04 « .008 (M)
LL(M)»-.04 « .008 (W.)
1
1
i
2
i
3
i
4
STAN
i
5
DARD
t
6
\
I
7 8
SAMPLE
i
9
SET
i
10
NO.
i
II ic
(M)
"" 11 \ ~ U U • 'v i K ^J L.
WITHIN
-------
IS)
14 .0
2.0 -
10.0 -
8.0 -
6.0 -
4.0 -
2.0 -
- 2.0 -
- 4.0 -
-6.0 -
LABORATORY A PRECISION CONTROL CHART
AMMONIA NITROGEN DISTILLATION METHOD
OCT. 10, 1968
RANGE * 1.5 to 6.5 mg/l NH3 - N
a =.15 A =.15
S
-------
References
(1) Eisenhart, C. "Realistic Evaluation of the Precision and Accuracy
of Instrument Calibration Systems", Journal of Research,
N.B.S., 67C (2): 161-187, April-June 1963.
(2) Kelly, W. D. "Quality Control", Unpublished presentation for the
National Center for Radiological Health.
(3) Shewhart, W. A. "Economic Control of Quality of Manufactured
Product", New York, D. Van Nostrand Co., 1931.
(4) Wernimont, G. "Use of Control Charts in the Analytical Laboratory",
Industrial and Engineering Chemistry, 18 (10): 587-592,
October 1946.
(5) Dixon, W. J. and Massey, F. T., Jr. "Introduction to Statistical
Analysis", McGraw Hill, 2nd. ed. 275-278, 1957.
(6) Wald, A. "Sequential Analysis", John Wiley and Sons, Inc., New
York, 1963.
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY'S
ANALYTICAL QUALITY CONTROL PROGRAM*
Revised by
Bobby G. Benefield**
INTRODUCTION
The Environmental Protection Agency (EPA) gathers water duality
data to determine compliance with water Quality standards, to provide
information for planning water resources development, to determine
the effectiveness of pollution abatement procedures, and to assist in
research and technical services activities. The sources of these data
are not only the EPA laboratories but other Federal, city, State, and
industry laboratories.
In a large measure the success of the pollution control program
rests upon the reliability of the information provided by the data
collection activities.
To insure the reliability of physical, chemical, and biological
data, the EPA's Division of Research has established the Analytical
Quality Control (AOC) Laboratory at 1014 Broadway, Cincinnati, Ohio.
* Originally entitled "The Federal Water Pollution Control Adminis-
tration's Analytical Duality Control Program" and written in 196^
by Mr. R. E. Crowe, then Chief, Chemistrv and Biology Section,
Technical Assistance, Technical Services Program, Robert S. Kerr
Water Research Center, Federal Water Pollution Control Administration,
U. S. Department of the Interior, Ada, Oklahoma. This revision was
presented bv Mr. Larrv J. Streck, Chemist, Chemical and Biological
Sciences Program, Office of Technical Programs, Environmental Protection
Agency, Robert S. Kerr Water Research Center, Ada, Oklahoma, during
Training Course No. 161.2, "Planning and Administrative Concepts of
Water Quality Surveys," March 22-26, 1971, at the Robert S. Kerr
Water Research Center, Ada, Oklahoma
** Analvtical Oualitv Control Regional Coordinator, Region VI, Environ-
mental Protection Agency, Robert S. Kerr Water Research Center, Ada,
Oklahoma.
-------
The program conducted by this laboratory is designed to assure the
validity and, where necessary, the legal defensibility of all water
quality information collected.
HISTORY
The Federal Water Pollution Control Administration (FWPCA), presently
EPA, recognized the need for an analytical quality control- program in
September 1966. Following this recognition, the first meeting on
analytical quality control in the FWPCA was held in Cincinnati, Ohio, in
January 1967. This meeting was attended by appropriate representatives
from FWPCA's nine (presently ten) regions.
The purpose was to bring together as a working group those FWPCA
personnel professionally and technically oriented and most knowledgeable
in analytical chemical methods and procedures used to identify, measure,
and characterize various types of water pollution. This group was
designated as the Committee on Methods Validation and Analytical Quality
Control. The Committee includes scientists who actively participate in
the preparation of Standard Methods for the Examination of Water and
Wastewater, American Public Health Association (APHA) and in subcommittee
and task group activities on Committee D-19 of the American Society for
Testing and Material (ASTM). In addition one of the scientists is
General Referee for Water, Subcommittee D of the Association of Official
Analytical Chemists.
At the Cincinnati meeting subcommittees were chosen to study the
then existing analytical chemical methods for investigating water quality
and to recommend the best of these methods for official designation by
the FWPCA. These subcommittees were further broken down into specific
-------
parameter groups which, for one reason or another, were related. A
Chairman was appointed for each of the subcommittees. The objectives
set by these subcommittees to be accomplished by the fall of 1967 were:
1. To select those parameters which would
be of use in the examination of water
quality.
2. To review the analytical methods avail-
able for analyzing these parameters.
3. To formulate a list of the best methods
available for immediate use.
Meeting these objectives represented the first major task to be accom-
plished by the newly organized Analytical Quality Control Section, which
at the time was a part of FWPCA's Division of Pollution Surveillance.
In October 1967, the Committee on Methods Validation and Analytical
Quality Control held its second meeting — again in Cincinnati. At this
time the initial task of the Analytical Quality Control Program was
complete. The product of this first pioneering step was the publication
and distribution of the FWPCA Official Interim Methods for Cnemioal
Analyses of Surface Waters, September 1968.
In addition to the methods selection and validation activitv, the
Analytical Quality Control Program was being re-organized as a Laboratory
of the FWPCA's Division of Research with the establishment of regional
coordinators throughout the United States to coordinate program activities
in each FWPCA region. Bringing the program to this point was a major step.
The AQC Laboratory is currently engaged in: (1) the selection and
validation of methods for biological and microbiological determinations
much the same as was done for chemical determinations (the biological
-------
methods manual will be available soon after January 1972); (2) intra-
laboratory quality control programs; (3) interlaboratory quality control
programs; and (4) publishing the third edition of the Methods for Chemical
Analysis of Water and Wastes, 1971.
ORGANIZATION
The Analytical Quality Control Program of the EPA is carried out
through an Analytical Quality Control Laboratory assisted by advisory
committees on methods selection, regional quality control coordinators,
and laboratory quality control officers. The organization and functions
of these groups are described below.
Analytical Quality Control Laboratory
The Analytical Quality Control Laboratory is composed of five
sections - Chemistry, Biology, Microbiology, Instrument Development,
and Methods and Performance Evaluation. The laboratory staff coordinates
the AQC Program, carries out methods development, conducts a continuing
reference sample service, and statistically evaluates laboratory
performance.
Regional Coordinators
To emphasize quality control participation at the regional level,
each Regional Director appoints a coordinator whose primary functions
are to implement the analytical quality control program in all EPA
laboratory activities within his region, and to assist or offer advice
to appropriate groups outside the EPA concerning any phase of analyt-
ical qualitv control in the laboratory. Through individual quality
control officers he provides leadership within the regional laboratory
components, insuring the usefulness of this data for all regional functions.
-------
In addition, the coordinator keeps the Regional Director advised on
analytical quality control activities in the laboratory under his
jurisdiction and informs the Analytical Quality Control Laboratory of
his region's needs in methods development and data validation. The
regional AQC coordinators and their respective regions are:
Regional Analytical Quality Control Coordinators
Francis T. Brezenski, AQC Coordinator
Environmental Protection Agency
Hudson-Delaware Basins Office
Edison, New Jersev 08817
Charles Jones, Jr., AQC Coordinator
Environmental Protection Agency
1140 River Road
Charlottesville, Virginia 22901
James H. Finger, AQC Coordinator
Environmental Protection Agency
Southeast Water Laboratory
College Station Road
Athens, Georgia 30601
LeRoy E. Scarce, AOC Coordinator
Environmental Protect icn Agencv
1819 West Pershine: Roari
Chicago, Ilj-
Robert I.. Booth. AOC Coordinator
F.nvin i-r.r.tul Froiecticn Agenrv
Analvtical Ounlit' Control Laboratory
1014 Broadwav
Cincinnati, Ohio 45202
Laboratorv
Control Officers
Harold G. Brown, AQC Coordinator
Environmental Protection Agencv
911 Walnut Street, Room 702
Kansas City, Missouri 64106
Bobby G. Benefield, AOC Coordinator
Environmental Protection Agency
Ada Facility, P. 0. Box 1198
Ada, Oklahoma 74820
Daniel F. Krawczyk, AOC Coordinator
Environmental Protection Agency
Pacific Northwest Water Laboratory
200 South 35th Street
Corvallis, Oregon 97330
Donald B. Mausshardt, AQC Coordinator
Environmental Protection Agency
Phelan Building, 760 Market Street
San Francisco, California 94102
John Tilstra, AQC Coordinator
Environmental Protection Ager.cv
Lincoln Tower Building, Suite 900
1860 Lincoln Street
Denver, Colorado 30203
This officer, -usually a senior member of the labcratorv staff, is
appointed bv the Laboratory Director and is responsible, through him,
to thf ?ec:cn.il 'Xi'ilitv Control Coordinator. He is concerned with the
analytical quality control of EPA laboratories within the region.
-------
RESPONSIBILITIES
National
The national responsibilities of the EPA's analytical quality
control program are primarily those of the Analytical Quality Control
Laboratory in Cincinnati, Ohio. These responsibilities are described
below.
Methods Research
Although analytical methods are available for most of the routine
measurements used in water pollution control, there is a continuing need
for improvement in sensitivity, precision, accuracy, and speed. Develop-
ment is required to take advantage of modern instrumentation in the water
laboratory. In microbiology, the use of new bacterial indicators of
pollution, including pathogens, creates a need for rapid identification
and counting procedures. Biological collection methods need to be
standardized to permit efficient interchange of data. The Analytical
Quality Control Laboratory devotes its research efforts to the improvement
of the routine tools of the trade; therefore, it has nationwide respon-
sibility for the guidance of a program to develop reliable analytical
methods for water and wastewater analyses.
Methods Selection
The AOC Laboratory provides the program for the selection of the
best available procedures in water and waste analyses. This includes
certification of the methods through an adequate testing program.
Through the publishing of EPA methods manuals, updated regularly, the
program insures the uniform application of analytical methods in all
laboratories of the EPA.
-------
Interlaboratory Quality Control
The Analytical Quality Control Laboratory is responsible for
maintaining a reference sample program for methods-verification and
laboratory performance evaluation of all EPA laboratories. This also
includes the validation of chosen procedures of existing or new develop-
mental methods of analysis.
Intralaboratory Quality Control
To maintain a high performance level in daily activities, everv
analytical laboratory must utilize a system of checks on the accuracy
and precision of reported results. While this is a part of the respon-
sibility of the analyst and his supervisor, the Analytical Quality
Control Laboratory is responsible for guidance in the development of
model quality control programs which can be incorporated into the
laboratory routine.
Regional
The regional responsibilities are essentially those of the regional
coordinator and the laboratory quality control officers of a particular
region. The regional coordinator is responsible for implementing the
nationwide program in the EPA regional laboratory and maintaining ap-
propriate relations with other federal agencies, with state and interstate
pollution control agencies, and with industry to encourage their use of
the EPA methods and their participation in the analytical quality control
effort. In addition, the regional coordinator is responsible for bringing
to the attention of the AQC Laboratory any special needs of his region
in analytical methodology and any analytical quality control problems that
occur.
-------
The laboratory quality control officer is responsible for carrying
out an intralaboratory quality control program within the EPA laboratory
in his region, assuring the use of certified methods by the laboratory
staffs and securing participation in regular check sample analyses.
EPA OFFICIAL METHODS
Missions assigned to EPA by the Water Quality Act of 1963 and the
Clean Water Restoration Act of 1966 created a need for methods capable
of developing water quality data to measure the effectiveness of the
Nation's water polluticn control programs. The methods must be uniform
throughout the Agency and based on sound, scientific investigations.
Further, thev must be available to all other elements of the water pol-
lution control field involved in, or affected by, water quality standards,
and must be acceptable as legally defensible in Federal and State
enforcement actions to abate water pollution.
The first edition of a methods manual was entitled the FWPCA Official
Interim Methods for C'nertioal Analyses of Surface Waters and was a major
step in this direction and represented the first product of the EPA's
Analytical Quality Control Program.
To acquaint you with the manual and its implications, it seems ap-
propriate to discuss it at this time. As I have said, this manual
represents the selections made by a committee of senior EPA laboratorv
personnel, working under the guidance of the Analytical Oualitv Control
Laboratory. The Committee consulted all the available literature, including
Standard Methods for tnc Examination of Water and Wasteuater^ ASTM Manual
of Industrial Water, and current technical journals. This manual in first
edition was limited in number, and thus was not available to the general
public. The second edition, entitled, F'vFJ.-'. Methods for Chemical Analyses
-------
of Water and *'aszes was published in July 1969 and was available for
all who desired copies. A third edition was published in 1971. It
is entitled Methods for* C'nemisal Analysis of '^'ater and Pastes and is
available upon request.
An analytical quality control manual is also available. It is
entitled Control cf C'-.emical Analyses in Water Pollution Laboratories.
This manual deals exclusively with quality control within the laboratory.
The 1971 methods and the laboratory qualitv control manual can be
obtained by sending a request to your regional coordinator. For those of
you who are in this region (Region VI) the address is
Robert S. Kerr Water Research Center
United States Environmental Protection Agencv
P. 0. Box 1198
Ada, Oklahoma 74820
Attention: Bobby G. Benefield
AQC Regional Coordinator
SUMMARY
In order that industry, state and interstate pollution control
agencies, the EPA and other Federal agencies can effectively relate
water quality data and feel secure that water quality data produced
from all the laboratories concerned are valid, the EPA organized an
analytical quality control program. The organization and responsibil-
ities of this program nationally and regionally have been discussed.
The success of this program depends upon the active participation
of managers, supervisors, and analysts in not only the EPA, but in
all groups concerned with characterizing and maintaining water quality.
Therefore, it is essential that the philosophy of quality control be
understood and accepted by all levels in anv laboratory organization.
-------
INTRODUCTION
In the "Quality Control of Chemical Analysis" section of this
manual, it was stated that one of the basic assumptions made in the
construction of control charts is that the spiked sample data or
duplicate data should be the products from an-"in-control" process.
This addenda offers a statistical method by which the validity
of this assumption mav be evaluated.
ELIMINATION OF OUTLIERS
If obviously large differences exist between matched pairs from
spiked or duplicate data and if an assignable cause for this difference
is not known, then an unbiased method for rejection of outliers must be
used. Two such methods are given below (1).
TEST 1: ESTIMATE OF od AVAILABLE
A statistic which can be used to detect outliers in either direction
(too large or too small's is q = W/S., where W is the range of the differ-
cn.es and S^ is an independent estimate of the population standard devia-
tion of t'T •! '. f f t'rencfs ('",-.'•
for -.entiles of the ->;i-pl;ns; distribution of q are given in Table 1.
If a significantly large \\ilue is obtained, it should not be used in
subsenuert cal rul "itions . A check should be made in an attempt to find
assignable raus£"= for Lue 1-irc^? value(s).
EXAMPLE I
I're1 .' - 1 .11 Control Chart
Laboratory: Laboratory A
Parameter Analyzed: Alkalinity as
M<=thnd:
Date: <.-•.-> . -' , •••-•
Da',-.-
-------
Results of Analyses of Duplicate Samples
(Mg/1 Alkalinity as CaC03)
Set No. Duplicate No. 1 Duplicate No. 2 Difference
1 96.0 100.0 -4.0
2 222.0 218.0 4.0
3 244.0 242.0 2.0
4 79.0 80.0 -1.0
5 524.0 526.0 -2.0
6 410.0 414.0 -4.0
7 118.0 118.0 0.0
8 70.0 70.0 0.0
9 50.0 50.0 0.0
10 297.0 303.0 -6.0
11 307.0 312.0 -5.0
12 296.0 303.0 -7.0
13 180.0 186.0 -6.0
14 211.0 214.0 -3.0
15 214.0 212.0 2.0
16 215.0 216.0 -].0
17 139.0 142.0 -3.0
18 122.0 124.0 -2.0
19 i:~.o 127.0 0.0
20 444.0 464.0 -20.0
21 109.0 100.0 0.0
23 R9.0 87.0 2.0
W = Range = d(max) - d(ir.in) = 4.0 - (-20.0) =24.0
Sjj = 3.7867 with 42 degrees of freedom*
q - W/Sd = 24/3.7869 = 6.3376
The 95 percentile value for the distribution of q = W/S^ with K = 23
and d.f. = 40 is 5..'.6 (Table 1). The computed q is greater than the tabu-
lated q, therefore, conclude that the di f f eri_-nce resulting from the dupli-
cate set /'20 is truly an outlier and should be eliminated from control
chart calculations. This procedure may be iterated for all suspected
-utlier -.
* Sj obtjint_c '""i" >r ', r -jo;v r ^nf <•-..(. -'T~ :•;;•". i r -;«• es v:;rh J'5, .i>"--.erved naivs.
-------
.
~D
O
"3
O
-a
"3
C
re
4-J
C
a)
-a
c
a.
o
•H
-------
TEST 2: ESTIMATE OF -• NOT AVAILABLE
d
A statistic which can be used to detect outliers when an estimate
of the population standard deviation of differences (a ) is not known
is described below.
The test proceeds as follows:
1. Arrange the data in ascending order.
2. If
3 < n - 7
8 < n $ 10
Urn. 13
14 < n r 25
compute r
10
compute r..-
compute r?1
compute r_~
n is the number of differences between matched pairs of spiked or
duplicate data. Compute r.. as follows:
r . .
io
21
if d is suspect
n
if d., is suspect
1
(d -
(d -
- V
3. Look up r qR for r.. as defined in Step 2 in Table 2.
4. If r. . - r no, reject the observation, otherwise, do not reject.
i j . ^ o
EXAMPLE II
Consider the data used in Example I.
I'".-.- ::•;-•• c-r of d'j-i 1 i cat e^ (n) lies between 25 and 14, therefore, we
:••"-.-u:f r, tl • f ^1J DI>-, :
-------
r22 = (d3 " V/(dn-2 " dl) = [("6) " (
= 14/22 = .6363,
which is greater than r no 0_ = .422. Therefore, we reject the
. y o, z j
suspected outlier.
This procedure may be iterated until all suspected outliers have
been checked.
TABLE 3. CRITERIA FOR REJECTION OF OUTLYING OBSERVATIONS
y
rio
ril
r21
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
22
19
20
21
22
23
24
' 25
.976
.846
.729
.644
.586
.631
.587
.551
.638
.605
.578
.602
.579
.559
.542
.527
.514
.502
.491
.481
.472
.464
.457
-------
TEST FOR "IN-CONTROL"
Once a useable set of data is at hand, it is necessary to compute
the mean difference and the standard deviation of the mean difference.
Since the theoretical difference between duplicates of the same material
is zero, Student's t distribution can be used to test the hypothesis that
the average difference of the population sampled differs significantly
from zero. If it does not, then the process is judged to be in control
and subsequent computations for constructing the control chart are con-
sidered valid.
The data in Table 1, with duplicate set #20 eliminated, will be used
for expository purposes. The t test is performed as follows:
d - average difference = -1.5652
S- = S, / /~K~ = 0.95869
d d
t = d / S-7 = -2.4339 with 23 degrees of freedom
The 95 percentile value for the t distribution with 22 degrees of
freedom (Table 3) is 2.074. The absolute value of the computed t is
greater than the tabulated t. Therefore, it is concluded that the mean
difference of the sampled population is significantly different from
zero.
A word of caution is noteworthy here in confusing the terms
"significantly different" and "meaningfully different." It is possible
to obtain a significant difference that is not meaningful. An example
would be a significant difference of .005 mg/1 when the accuracy of the
measuring procedure is only, say, ± .05 mg/1. In this case, the data
would be judged to be in control, and the control chart constructed from
the data would be considered valid.
-------
TABLE 3. Values of Student's t
Probability of a larger value of a t » .05
d.f.
1
2
3
4
5
6
7
8
o
10
11
12
13
i<«
15
16
17
16
19
20
21
22
23
24
25
26
27
28
29
30
40
60
120
-
n -
t
12.706
4.303
3.182
2.776
2.571
2.447
2.365
2.306
2.262
2.228
2.201
2.179
2.160
2.145
2.131
2.120
2.110
2.101
2.093
2.086
2.080
2.074
2.069
2.064
2.060
2.056
2.052
2.048
2.045
2.042
2.021
2.000
1.960
1.960
0.025
-------
BIBLIOGRAPHY
1 Dixon, W. J. and Massey, F. T., Jr. "Introduction to Statistical
Analysis", McGraw-Hill, Second Edition
-------
COMPUTER APPROACH
TO
QUALITY CONTROL PROCEDURES
INTRODUCTION
The measure of effectiveness of any procedure requiring
mathematical manipulation of numbers is most often inversely
proportional to the amount of hand calculations required. In an
effort to minimize the mathematical involvement of the laboratory
scientist when using these quality control procedures, a computer
prccrarr. has been developed and refined in such a way as to give all
pertinent information in a well formated, easy to use and store
printout.
The utility of this program, of course, depends upon the
availability of some form of data processing equipment. For those
who have a computer available for their use, the following documentation
package is provided on a Fortran IV program written for an 8K IBM 1130
with a Disk Monitor System. This program could be easily modified for
any computer system having a Fortran Compiler.
-------
In using spiked or standard samples to check accuracy, a significant
t value may result due to a consistent over or under reporting of concen-
trations. This is bias inherent in the procedure. Efforts should be
made to ascertain the cause for this discrepancy and remove it if possible.
If it can not be eliminated, past experience on the part of the analyst
must suffice in determining if this difference is meaningful.
-------
SIGMA QUALITY CONTROL
A. ABSTRACT
The Sigma Quality Control Chart program has been designed to
calculate basic descriptive statistics and the control line equations
necessary for constructing cumulative-sum quality control charts.
The input is in the form of duplicate or paired standard and
observed values. The following output is provided for each data set.
1. Sample identification
2. Original data
3. Basic Descriptive Statistics
a. average difference (DEAR)
D. standard deviation of the average difference (SDBAR)
c. computed Student's "t" value (T)
d. ALPHA
e. BETA
f. DELTA
g. variance of the differences
h. standard deviation of the differences
i. sum of differences
j. sum of squares
K. maximum allowable variance (S(1}SQUARED)
1. ir.imimum allowable variance (S(0)SQUARED)
4. Equations for upper limit line and lower limit line evaluated
for M = 6 and 10.
B. METHOD OF SOLUTION
7 .£ c o-r •..: ~ r -rives frr values according to the following set
-------
d = DEAR = [Z (X. - Y.)]/N, = (I d.)/N
2.
3.
A.
5.
6.
where N = Number of pairs.
S2 = [I d.2 - (I d.)2/N]/(N - 1)
d i i i i
C _ ., C^
bd v bd
S- = SDBAR = S ./•N~
d d
S2 = (1 - A)2 • S2
o d
Sj = (1 + A)2 • S2
UL(M) =
i-e
(M)
8. LL(M) =
2 log
6
1-a
loge
S2
1
C £-
O
1 _ J_ ' 1 _ 1
S- ~ c2 c2 ~ c2
0 1 0 1
C.
FORMATS
1
Control Card
cc
1- 3
A- 6
7- 9
10-11
13-33
35-55
57
5^
61
63
16-80
ITEM
Number of pairs of data
Alpha
Beta
Delta
Parameter name
Description of method
X if wet lab, blank otherwise
X if instrument lab, blank otherwise
X if Precision Control Chart, blank otherwise
X if Accuracy Control Chart
Range data covers
-------
2. Data Cards
cc ITEM
1- 7* value for first duplicatet or standard?
8-14* value for second duplicate'" or observed1
* Decimal point must be punched.
~ for duplicate data on Precision Control Charts.
f for standard and observed values on Accuracy Control Charts.
3. Last Card ITEM
cc
1-80 9's
D. OPERATING PROCEDURES
1. Load data into hopper of 1442 and press START.
2. Press IMMEDIATE STOP, RESET, and PROGRAM LOAD on the CPU.
3. Ready printer.
4. When the kevboard select light comes on, type in a six digit
date, i.e., 071369.
E. DECK KEY
¥¥9999999999
/Control Card
/Data Deck
/Control Card
ta Deck
/
>) Control Cards
r &
Data Deck
"EXTENDED PRECIS!
/"
IOCS
/Cold Start 1
DN
-------
F. PROGRAV 1:^7 I\C,
2 A 0 F. 1
/ / J C - '•' • f >
L C "- 0 c ! V E C A K T S >~
JjJv 222i
CART AVAIL PHY DRlVt
^222 0000
A C T j A _ r < C C N F I G « K
// POP
* I oc s < c A « r , T v P t A r< HER. KEYBOARD, 113? P R i N T E R • D i so
* L i s ' s ~ 'j V c E n ? o r- K A •-'
*- x T1"".'";1 ": • •'-' r c ; : : r •.
•;:'•'' ,s if/. ,,L 1 2 1 .XL ( i ) »•'.( ^ ) .PAR ( 2. ) »Ir AC ( 6 « ^ ; i I :AY ,•.«(), j YP
•3? crCp-'«,T , - ; ; j
fcfi^ .VR ;T» ; 3, t- i j
«C1 i- OR^AT ( I'-l )
C cr%T~C,L CA^D \ IS THE fjU^flER OF CARDS A IS ALPHA B IS BETA
- -~ATi 2 » 1 > '>i»A ,POtPA3 »'«FTH »WET » !NST»PREC»ACCtRANG£
I F ( A - 9 . (: 9 ) 4 4 4 , f 1 3
,-;~; -- ; 3 ,f:) IDAV.^C. IY <
5C ^C^'-'A* , 1 K , ' ^CBERT S. KEPR w'ATER RESEARCH CENTER ' » 26X » I 2 » ' - ' . I 2 « ' - '
*» 12 »/ /<'»1X» 'SIG'-'A QUALITY CONTROL CHART INFO.1)
,-. - i TE ! 3 , 4CC • >PAR , I'~TH , * E T i I NS T . P R EC • ACC • RANGE
-;:-'AT (//, . p.-ARAMETr-5—S21A1,' METHOD—' . 2 1 Al . / / »IX , ' *'ET—'.Alt'
* INS*.— ',A;,' PREC.— 'tAl.' ACC.— '»Ali' KANGE S15A1 )
- C. R '•' A ^ ; / , ,:- ^ , ' v • , i 2 X , ' Y ' , 12 X , ' C ' )
4C
E SU"' OF ThE SQUARES
v A - ; ] t , -' i 2 . 4 i 1 x » r 1 2 . 4 , 1 X » F 1 2
-K VALANCE AND STANDARD DEVlATIOiM
- .-. A ,-, = 5 j v / x \
£ ~ ^ A •- = i T ; , / x ' . * * ,
T = .: - A K / 5 7 u /•. r
, '- 4 , ,
« 9 C V ) .0 R A R » S ^ 5 A R » T » NM 1
/, ' -,f-A^ ' , _- IX , ( = ',F16.7./.1X
P 1 -, . - , ; x , ' v. I T H ' » I 3 t ' D . F .
30X
F16.7,/,iA
-------
PA3E 2
pl= ( ( 1-3) /A!
F2=51/SO
Al = (2-w-ALCGlFl ) )/< ( 1/SOJ-t I/SI) )
DO 11 L= 1.2 " "
IF (L.-2 ! 12.13.13 " " __
12 M(l)=6 .-.-.-
GO TO 14 " " "" ......... "
i3 M(2!=10 " " "" "
14 61=(ALOG(F2 )/( ( l/SO)-( I/SI ) ) }" '
B2=B1*M(L) "~"
_
A2 =(2*ALOG(F3) )/( ( l/50)-( I/SI ) ) '" ..... "" " ......... ""
11 XL(L)=A2 +B2 ' ' '"
WRITE<3»A) A,B»D
4 FOR^ATUX, 'ALPHA' » 30X , ' = « ,7X , FA . 2 . / . IX » 'BETA' • 3 IX . "« = ' » 7X »F4.2 . / . IX
*.'DELTA',30X»'=',7X.F4.2)
V.'RITE !3»22 )VAR»STD " " " ..... "" ' "" ..... "
22 FCR'-'AT< IX'VARIANCE OF THE D I FFERENCES '". 8X ,'=''". ' F16 . 7 . / , IX t"' STAND
*ARD DEVIATION OF D I FFERENCES ' > 2X , ' = ' » F16.7) ......
A'RI'E (3 .21 ) SUN'iSUf^SQ
21 ^OR'-'ATI IX, 'SU'-' OF D I FFERENCES '» 1 7X »' = '»" F16.7 »/»lX»'SUM OF SOUA
*RE£ ' »21X» ' = » » F16.7)
K^ITE ! 3 »6 ) SO. SI
6 FGR'-'AT I //. ' SIO) SQUARED* ' . F16 . 7 . 3X , ' S ( 1 ) SQUARED= ' . F16 . 7 )
V. R ! T E ( 3 , 7 )
7 FOR'-'AT(//,20X » ' A ' . 2CX i ' B ' » 9 X.'M')
V, R I T E ; 3 , B 1
6 -CPVAT ! 1X.7K '-' ) ) •- : '
C02G <=1 »2
WRITE (3»9 )A1 iBl »M«J »UL «)
9 FO-VAT ( IX. 'UL(^') = ' ,2X. F15.6»4X» ' -t- ' »2X»F15.6» ' (M) ' »2X» I2»3X. ' = '"»F
£15.6)
20 CONTINUE ' " " ..... " " " ..........
D025K = 1 ,2 -- - - ......... ---
WRITE (3, 10) A2 . Bl . M « ) , XL t < ) "" " ....... " ...... .......
10 FOR '-'AT! IX. 'LHV ) = ' »1X,F16.6»AX» '-*•' »2X»F15i6» ' (M) ' »2X»I2»3Xi ' = ' »F15
*.&)
,25 CONTINUE " ....... ' " " ^ "" .......
I/, = ;TE 1 1 «eoo) ......... ~_~"^ ~ ...... ' ......
SCC "DF/.-AT ! IX, 'END OF JOB1) ..... " """ ......... ~""~ .....
SIC CALL EXIT
END
FEATURES SUPPORTED
EXTENDED PRECISION
IOCS
CORE REQUIREMENTS FOR
COMMON C VARIABLES 3CA PROGRAM " " 976
CC'-'"lLAT ION " ' " "" "
-------
SAMPLE
X X X X 1-l.MG/L
------- |