This program is best viewed with screen

settings: 600-800 pixels. Introduction

In-house validation of
For question about this program please
contact: P.A. Zandbelt
URDV-Quality Systems
Peter.Zandbelt@Unilever.com

analytical methods
Procedures & Calculation sheets
versie 2, 21-6-2004

A full text version of this program can be found in UMA method 1009, Validating an analytical method

VALIDATION OF THIS PROGRAM

If you have installed this program for the first time, testing must be performed prior to using the
program on real data. Please go to the validation part of this program by pressing the button
"Validation of this Excel program".

Warning: Most of the equations used in these sheets assume a full test as proposed by this
validation program. The equations are mostly not valid in deviating situations. To avoid any
obscurity there will be no output in several cases when not all results have been filled in.

This program contains instructions and calculation sheets for in-house validation of quantitative analytical methods.
In case of an in-house validation a part of a complete validation (which is a validation carried out when a new method
is developed and includes 6 different studies) is repeated to check if an analytical method is functioning properly in
a specific laboratory. In-house validation first of all includes the validation strategy when an official analytical method
or a method that is developed and validated elsewere is implemented for the first time in a laboratory. Furthermore it is
a validation strategy when the operation of a method is changed by using another analyst and/or instrument or when
the analysis prescription is modified by extending the concentration area, using a new matrix or changing other parts
of the method.
Four validation studies are described. The upper two are mandatory for an in-house validation. The lower two are only
needed in specific situations. If you do not know which studies are needed in your specific situation the scheme at
the bottom of this sheet might help you.

The validation studies described here are not new, and are extensively described in literature. What is new is that this
program provides you with fully automatic calculation sheets that circumvent the need for a profound knowledge of
statistics. More information about these studies is available in UMA method 1009 (Validating an analytical method)

Please click on one of the buttons to go to the sheet that describes the specific study

TRUENESS STUDY

PRECISION STUDY TO DETERMINE

REPEATABILITY AND WITHIN LABORATORY REPRODUCIBILITY
LIMIT OF DETECTION AND QUANTIFICATION
(for trace analysis only)
LINEAR WORKING RANGE (when extending the scope of the method
only, e.g. concentration area or matrix)

The scheme below can help you to decide which studies are needed for your in-house validation.
The 6 blue boxes contain all the different validation studies possible when a complete validation
is needed. (validation of a new developed method). The options outlined in red and green are
specific for in-house validation. This program contains procedures and automatic calculation
sheets for these 4 studies.

More information about the procedures used in this program can be found in:

Vovol course method validation, B.G.M. Vandeginste and L.C.M. Buydens

Standard Operating Procedure 548/01, Validation of Analytical methods:

Guidelines for the conduct of a validation study, P.C. van Bruggen,
H.C.M. van der Knaap, F.J. Slikkerveer and B.G.M. Vandeginste

Statistics for Analytical Chemistry, 3rd edition, J.C. Miller and J.N. Miller
ISBN 0-13-030990-7

UMA method 1009, Validating an analytical method, Peter van Bruggen,

Peter Zandbelt
TRUENESS STUDY
The trueness is the closeness of agreement between the average value obtained from a large series of test results
and an accepted reference value. The measure of trueness is usually expressed in terms of bias (ISO 5725-1:1994).
Another term associated with trueness is systematic error (NEN3114).

Statistical model:

X = Xw + δ + ε
Measured True systematic random
value value deviation error

Trueness Precision

Accuracy

The bias contains two contributions. The method bias that is inherent in the analytical method and the laboratory bias
that is attributed to a particular laboratory. Method bias refers to the deviation between the true value and the value
obtained with the method. It is well possible however that in the absence of a method bias, a particular laboratory
on the average finds a deviating result. This is the laboratory bias. When a trueness study is performed in a single
laboratory the method-bias and laboratory-bias are confounded and cannot be distinguished. In many cases the
so-called true value is not known and instead an accepted or consensus value is used.

The bias may be a constant deviation (off-set) from the true value. It may also be that the bias is proportional to the
concentration or to the amount (proportional bias).

The procedure which should be followed to carry out a trueness study is highly determined by the availability
of certified reference materials (CRM), test samples, standards or blank samples spiked with the analyte.
The type of bias which can be determined and the procedure which should be followed are summarised in the table below.

Please click on one of the buttons to continue with the desired study.

Possible trueness studies for in-house validation

Materials and Methods What is determined Click to go further
A Certified Reference Material (CRM) A bias at the concentration level of the CRM
Blank samples spiked with standards or a CRM An off-set and proportional bias for a concentration range
Test samples spiked with standards or a CRM The presence of a proportional bias. Recovery
CRM spiked with standards An off-set and proportional bias for a concentration range. Recovery
Test samples and a ref method, 1 conc. level A bias between both methods
Test samples and a ref method, various conc. levels An off-set and proportional bias between both methods

Explanation of terms used

Certified reference materials
Certified reference materials are materials of which the true value, accepted or consensus value is certified within certain limits.
For the purposes of this program, we assume that the precision of the stated composition is much better than the precision of the
analytical method that is to be validated.
When a CRM is used to determine the trueness of an analytical method two conditions must be fulfilled:
The CRM must fall within the scope of the analytical method, i.e. the CRM must be very similar to the samples that
are commonly analysed with the analytical method.
The concentration of the component to be determined must fall within the (linear) range of the analytical method.
If the concentration is too high, it is sometimes possible to dilute the CRM either by appropriate dissolution or by
adding a sample blank.

Blank sample
A blank sample is a sample which contains all constituents in roughly the same proportions as the test sample but none
of the analyte.

Standard
A standard is a sample which contains a known amount of the analyte but no matrix.
1. Trueness study with a certified reference material
A bias at the concentration level of the CRM can be determined with this study.

Procedure
1 Take 6 test portions from the Certified Reference Material (CRM).
2 Analyse each test portion in singular with the method that is to be validated.
Each test portion must undergo the complete analytical procedure including sample preparation.
When several analysts perform the analyses in practice, it is recommended either to determine the specific bias per
analyst/instrument combination or to divide the described tests among the analysts/instrument combinations.
3 Enter the results in the following table (blue fields). It is important that you fill in all fields. The conclusion is
generated automatically.

CRM Result Concentration of Calculations Conclusion

portion CRM
1 average no statistical significant bias
2 standdev
3 test value Note: This conclusion is only valid at
4 table value (n=6) 2.571 the concentration level of the CRM
5
6

4 Conclusion
If a bias is detected it is recommended that the analytical method is subjected to a careful inspection to
improve the results. If an improvement cannot be acquired the test results can be used as a correction,
however with constraint. In the absence of further, more extensive tests, correction is allowed when:
The observed bias is modest, no more than 10% of the component concentration.
The concentration of the sought component in the test samples is close to the concentration in the CRM.
2. Trueness study with a blank sample spiked with standards or CRM
An off-set and proportional bias for a concentration range can be detected with this study. Calculations are
performed by regression analysis.

Procedure
1 Divide a blank sample in 4 roughly equal portions.
2 1 portion is taken as such (=blank)
3 3 test portions are spiked with different known amount of the component to be analysed (via either a CRM or standard).
The added amount are respectively low, intermediate and high (tested working range).
4 Analyse the 4 test portions (so including the blank) in duplicate with the method that is to be validated.
Each test portion must undergo the complete analytical procedure including sample preparation.
The samples must be analysed in random order.
5 Enter the results of the 8 analysis in the table below (blue fields). The graphical output and regression analysis are
generated automatically.

test known amount analysis results 1.00 x + NaN

f(x) = NaN Calculations by Least Squares Regression
portion of added (duplicate) 0.80

analysis results
component 1 2 0.60
x y1 y2 yaverow xi-xave (xi-xave)2 yi-yave (yi-yave)2 (xi-xave)(yi-yave) x2 ycal (y1-ycal)2 (y2-ycal)2 sum
blank 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
0.40
low 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
0.20
intermediate 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
high 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
00 10 20 30 40 50 60 70 80 90 00 total 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0!
0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1.
known amount of component average 0.00 0.00

r: #DIV/0! sy/x: #DIV/0!

6 Regression analysis: results & conclusion b:(slope) #DIV/0! sb: #DIV/0!
a:(intercept) #DIV/0! sa: #DIV/0!
Important Details from regression analysis GENERAL CONCLUSION
intercept #DIV/0! #DIV/0! CIb/2 #DIV/0!
95% Confidence interval intercept/2 #DIV/0! #DIV/0! CIa/2 #DIV/0!
slope #DIV/0! #DIV/0!
95% Confidence interval slope/2 #DIV/0! #DIV/0! These tables are derived from: Statistics for Analytical Chemistry
3rd edition
7 If a bias is detected it is recommended that the analytical method is subjected to a careful inspection to J.C. Miller and J.N. Miller
improve the results.

If a statistically significant off-set or proportional bias is not detected, than this does not mean that the true
bias is zero.
It is therefore recommended to report the 95% confidence intervals of the intercept and the slope and the
maximum values of the bias as follows:

Intercept + CIintercept/2 #DIV/0! + #DIV/0!

Slope + CIslope/2 #DIV/0! + #DIV/0!

The maximum off-set bias, Max(Biasoffset) = IinterceptI+CIintercept/2 #DIV/0!

The maximum proportional bias, Max(Biasprop) = Islope-1I+CIslope/2 #DIV/0!

The maximum proportional bias can be expressed as a percentage:
The maximum proportional bias %, = 100 * Max(Biasprop) #DIV/0! %
3. Trueness study with a test sample spiked with standards or CRM
The presence of a proportional bias and the recovery of analyte at the 3 concentration levels can be detected
with this study. The possible presence of an off-set bias cannot be detected by this type of set-up.
Calculations are performed by regression analysis.

Procedure
1 Divide the test sample in 4 roughly equal portions.
2 1 portion is taken as such (unspiked test sample).
3 3 test portions are spiked with different known amount of the component to be analysed (either a CRM or standard).
The added amount are respectively low, intermediate and high (tested working range).
4 Analyse the 4 test portions (so including the blank) in duplicate with the method that is to be validated.
Each test portion must undergo the complete analytical procedure including sample preparation.
The samples must be analysed in random order.
5 Enter the results of the 8 analysis in the table below (blue fields). The graphical output and regression analysis is
generated automatically.

test known amount analysis results 1.00 x + NaN

f(x) = NaN Calculations by Least Squares Regression
portion of added duplicates 0.90
0.80
component 1 2 0.70
x y1 y2 yaverow xi-xave (xi-xave)2 yi-yave (yi-yave)2 (xi-xave)(yi-yave) x2 ycal (y1-ycal)2 (y2-ycal)2 sum

analysis results
unspiked test sample 0.60 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
low spiked test sample 0.50 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
0.40
intermediate spiked test sample 0.30
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
high spiked test sample 0.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
0.10 total 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0!
0.00
0.000.100.200.300.400.500.600.700.800.901.00
average 0.00 0.00
known amount of component
r: #DIV/0! sy/x: #DIV/0!
6 Regression analysis results & conclusion b:(slope) #DIV/0! sb: #DIV/0!
a:(intercept) #DIV/0! sa: #DIV/0!
Important Details from regression analysis GENERAL CONCLUSION
intercept #DIV/0! #DIV/0! CIb/2 #DIV/0!
slope #DIV/0! CIa/2 #DIV/0!
95% Confidence interval slope/2 #DIV/0! Note: An off-set bias cannot be detected by this type of set-up
These tables are derived from: Statistics for Analytical Chemistry
7 If a bias is detected it is recommended that the analytical method is subjected to a careful inspection to 3rd edition
improve the results. J.C. Miller and J.N. Miller

If a statistically significant proportional bias is not detected, than this does not mean that the true
bias is zero.
It is therefore recommended to report the 95% confidence intervals of the slope and the maximum
value of the proportional bias as follows:

Intercept #DIV/0!

Slope + CIslope/2 #DIV/0! + #DIV/0!

The maximum proportional bias, Max(Biasprop) = Islope-1I+CIslope/2 #DIV/0!

The maximum proportional bias can be expressed as a percentage:
The maximum proportional bias %, = 100 * Max(Biasprop) #DIV/0! %

8 Next to this the recovery of analyte at the three concentrations can be calculated
(Note: a proportional bias may cause a poor recovery at other concentration levels)
Recovery at low concentration (%) #DIV/0! %
Recovery at intermediate concentration (%) #DIV/0! %
Recovery at high concentration (%) #DIV/0! %
4. Trueness study with a CRM spiked with standards
An off-set and proportional bias for a concentration range can be detected with this study. Next to this the
recovery of analyte at the 3 concentration levels can be detected. Calculations are performed by regression analysis.
Note: spiking of a CRM with standards is only sensible if the concentration of the analyte in the CRM is below the range
of the method.

Procedure
1 Take 4 test portions from the Certified material (CRM).
2 1 portion is spiked with blank standard (or taken as such if no blank standard is available).
3 3 test portions are spiked with different known amount of the component to be analysed ( in the form of a standard).
The added amount are respectively low, intermediate and high (tested working range).
4 Analyse the 4 test portions (so including the blank) in duplicate with the method that is to be validated.
Each test portion must undergo the complete analytical procedure including sample preparation.
The samples must be analysed in random order.
5 Enter the results of the 8 analysis in the table below (blue fields). The graphical output and regression analysis
is generated automatically.

test known amount analysis results 1.00 x + NaN

f(x) = NaN Analysis via Least Squares Regression
portion of added (duplicate) 0.90
0.80
component 1 2 0.70
x y1 y2 yaverow xi-xave (xi-xave)2 yi-yave (yi-yave)2 (xi-xave)(yi-yave) x2 ycal (y1-ycal)2 (y2-ycal)2 sum

analysis results
CRM (+blank) 0.60 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
CRM + low st. 0.50 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
0.40
CRM + intermediate st. 0.30
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
CRM + high st. 0.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0! #DIV/0! #DIV/0! #DIV/0!
0.10 total 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 #DIV/0!
0.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
average 0.00 0.00
known amount of component
r: #DIV/0! sy/x: #DIV/0!
b:(slope) #DIV/0! sb: #DIV/0!
6 Regression analysis results & conclusion a:(intercept) #DIV/0! sa: #DIV/0!

Important Details from regression analysis GENERAL CONCLUSION CIb/2 #DIV/0!

intercept #DIV/0! #DIV/0! CIa/2 #DIV/0!
95% Confidence interval intercept/2 #DIV/0! #DIV/0!
slope #DIV/0! #DIV/0! These tables are derived from: Statistics for Analytical Chemistry
95% Confidence interval slope/2 #DIV/0! #DIV/0! 3rd edition
J.C. Miller and J.N. Miller
7 If a bias is detected it is recommended that the analytical method is subjected to a careful inspection to
improve the results.

If a statistically significant off-set or proportional bias is not detected, than this does not mean that the true
bias is zero.
It is therefore recommended to report the 95% confidence intervals of the intercept and the slope and the
maximum values of the bias as follows:

Intercept + CIintercept /2 #DIV/0! + #DIV/0!

Slope + CIslope/2 #DIV/0! + #DIV/0!

The maximum off-set bias, Max(Biasoffset) = IinterceptI+CIintercept/2 #DIV/0!

The maximum proportional bias, Max(Biasprop) = Islope-1I+CIslope/2 #DIV/0!

The maximum proportional bias can be expressed as a percentage:
The maximum proportional bias %, = 100 * Max(Biasprop) #DIV/0! %

8 Next to this the recovery of analyte at the three concentrations can be calculated
(Note: a proportional bias may cause a poor recovery at other concentration levels)
Recovery at low concentration (%) #DIV/0! %
Recovery at intermediate concentration (%) #DIV/0! %
Recovery at high concentration (%) #DIV/0! %
5. Trueness study by comparison with a reference method on one
concentration level
A bias between two methods can be detected with this study.

Procedure
1 Take six test portions of a representative test sample (it is not necessary that the concentration of the
analyte is known exactly).
2 Analyse each test portion in singular with the method that is to be validated, and with a reference method.
Each test portion must undergo the complete analytical procedure including sample preparation.
The trueness of the reference method must of course be known exactly.
3 Enter the results in the following table (blue fields). It is important that you fill in all fields. The conclusion is
generated automatically.

Test Result with Result with Calculations Conclusion

portion unvalidated reference
method method
1 unvalidated average
2 method standdev
3 reference average Note: This conclusion is only valid at
4 method standdev the concentration level of the
5 bias test sample
6 standdev bias
critical distance

4 If a bias is detected it is recommended that the analytical method is subjected to a careful inspection to
improve the results.

If a statistically significant bias is not detected, than this does not mean that the true bias is zero.
It is therefore recommended to report the maximum value of the bias (Biasmax) as follows:

Biasmax = BIAS + critical distance #VALUE!

6. Trueness study by comparison with a reference method at various
concentration levels
An off-set and proportional bias between two methods can be detected with this study. Calculations are
performed by regression analysis.

Procedure
1 Take six representative test samples with a concentration evenly distributed over the concentration range
to be investigated (it is not necessary that the concentrations of the analyte are known exactly, they must be
known approximately).
2 Analyse each test sample in singular with the method that is to be validated, and a reference method.
Each test sample must undergo the complete analytical procedure including sample preparation.
3 The repeatability of both methods (expressed as a standard deviation) should be known exactly. This can
be obtained via a precision study. Please enter these figures in the table below:

method standard
deviation
to be validated
reference

4 Enter the results in the following table (blue fields). It is important that you fill in all fields. The conclusion is
generated automatically.

Test Concen- Result Result calculations

portion tration with with difference mean standard critical s-ratio r Slope + CIslope/2 Intercept Analysis via Least Squares Regression
reference unvalidated (d) difference deviation difference (λ) (LSR)

method method (dave) (sd) (CD) x y xi-xave (xi-xave)2 yi-yave (yi-yave)2 (xi-xave)(yi-yave)
1 low 0.00 0 0 0 ### #DIV/0! ### + ### #DIV/0! 0.00 0.00 0.00 0.00 0.00 0.00 0
2 0.00 12 x + NaN
f(x) = NaN 0.00 0.00 0.00 0.00 0.00 0.00 0
R² = 0
3 0.00 10 0.00 0.00 0.00 0.00 0.00 0.00 0
unvalidated method

4 0.00 8
0.00 0.00 0.00 0.00 0.00 0.00 0
6
5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0
4
6 high 0.00 2
0.00 0.00 0.00 0.00 0.00 0.00 0
average #DIV/0! #DIV/0! 0 total 0.00 0.00 0.00 0.00 0.00 0.00 0
standard deviation #DIV/0! #DIV/0! C1 0 #DIV/0!
2 4 slope:
6 8 10 ###12 average 0.00 0.00
reference method
xave/sx yave/sy C2 #DIV/0!
r: #DIV/0!

CONCLUSION This table is derived from: Statistics for Analytical Chemistry

#DIV/0! 3rd edition
#DIV/0! J.C. Miller and J.N. Miller
#DIV/0!

5 If a bias is detected it is recommended that the analytical method is subjected to a careful inspection to
improve the results.

In the absence of a significant proportional bias, we recommend to report the 95% confidence interval
and the maximum value of the absolute bias as follows:

Biasoffset = dave + CD 0.0000 + 0.0000

The maximum off-set bias:
Max(Biasoffset)= IdaveI + CD 0.0000

In the presence of a significant proportional bias, we report

Biasprop = slope + CIslope/2 #DIV/0! + #DIV/0!

The maximum proportional bias:

Max(Biasprop) = Islope-1I+CIslope/2 #DIV/0!
The maximum proportional bias
can be expressed as a percentage:
Max(Biasprop) % = 100 * Max(Biasprop) #DIV/0! %
PRECISION STUDY TO DETERMINE REPEATABILITY AND WITHIN
LABORATORY REPRODUCIBILITY
The precision is the closeness of agreement between independent test results obtained under stipulated conditions.
Precision depends only on the distribution of random errors. The measure of precision is usually expressed in terms of
imprecision, and computed as a standard deviation of the test results. Less precision is reflected to a larger standard
deviation (ISO 5725-1:1994).

Statistical model:

X = Xw + δ + ε
Measured True systematic random
value value deviation error

Trueness Precision

Accuracy

Two important measures of precision are repeatability and reproducibility. The repeatability (expressed as r) represents
the smallest possible variability between testresults. The repeatability is defined as the maximum difference between
two analyses, that may be expected when they are carried out within a short time in a single laboratory by the same
operator using the same instrumentation. In contrast with this the reproducibility (expressed as R) represents the
highest possible variability between test results. The reproducibility, or more complete, the between-laboratory
reproducibility is defined as the maximum difference between two analyses, that may be expected when they are
performed by two different laboratories, and hence also different operators and instrumentation.
The reproducibility contains the largest amount of sources of variability, including those caused by the fact that different
laboratories use the method. If one implements an official method in a laboratory one is of course only interested in
the variability within the laboratory. For this purpose it is necessary to determine a within laboratory reproducibility.
This is a precision parameter that lies between the repeatability and the reproducibility.
Depending on the local situation several sources of variability can be of importance and taken in the precision study to
determine the within laboratory reproducibility. The factor "time" will always be of importance because the method will
be used on different days. Next to this if a method within the laboratory is used by different analists and/or different
instruments then the factors "operator" and "equipment" can also be incorporated in the precision study.

ISO uses the terms time-different intermediate precision standard deviation (S I(T)), operator-different intermediate
precision standard deviation (SI(O)), equipment-different intermediate precision standard deviation (S I(E)) and all
combinations (SI(TOE)).

The repeatability can be determined in a separate study but can also be obtained from the precision study in which
the within laboratory reproducibility is determined.

Please choose the appropriate study for your specific situation from the table below.
Next: click on one of the buttons to continue with the desired study.

Local situation Type of precision study click to go further

Separate repeatability study

The method is always carried out by the Precision study in which only
same analyst and on the same instrument the factor time is changed

The method is always carried out by 2 Precision study in which the

analysts on one and the same factors time & operator are changed
instrument
The method is always carried out by 2 Precision study in which the factors
analysts. Each analyst uses his own time & operator are changed. Operator and
instrument. instrument are a fixed combination.
The method is always carried out by one Precision study in which the factors
analyst and on 2 instruments time & instrument are changed.

The number of analysts that use the Precision study in which the
method times the number of instruments factors instrument & operator are
used by all analysts is larger than 2 (more varied over time.
than 2 analyst/instrument combinations)
1. Repeatability study
Procedure
1 Take 4 samples with concentrations distributed over the working range.
2 Analyse these samples in 5-fold on the same day. If it is impossible to perform al analyses on the same
day, the number of samples may be reduced but should never be lower than 3 levels in duplicate.
Each test portion must undergo the complete analytical procedure including sample preparation.
3 Enter the results in the following table (blue fields). The repeatability is calculated automatically.

Please note that the repeatability values (r) are calculated per concentration level separately and pooled.
This pooled, overall repeatability is only valid if the plot of the concentration vs the standard deviation does not reveal a
concentration dependency (see plot below the table).

concentration Concentration testnr result Mean standard repeatability (r) pooled

level (approx. known) deviation per conc. level repeatability
1
2 Overall repeatability
1 3 conc level n n-1 var (n-1)*var
4 1
5 2
1 3
2 4
2 3 sum 0 0 0
4
5 s 2
pooled #DIV/0!
1 roverall #DIV/0!
2 (df=0)*
3 3
4
5
1
2
4 3
4
5

results vs concentration level

12

10

6
SD

0
0 2 4 6 8 10 12
concentration

If the plot shows a linear relation between the SD (standard deviation) and the concentration, report both the relation and the
repeatability values per concentration level. If the plot does not show a relation between the SD and the concentration level the
pooled repeatability can be reported.
2. Precision study to determine repeatability and within
laboratory reproducibility
Precision study in which only the factor time is changed
(for 1 analyst using 1 instrument)
Procedure
1 Take 3 samples with concentrations distributed over the working range.
2 Analyse these samples in duplicate on the same day.
3 Repeat these analyses on 4 different days giving in total 24 analytical results.
4 Enter the results in the following table (blue fields). It is important that you fill in all fields.
Set-up of this test:

analysis nr day concentration duplicate result Concentration high

Results Concentration High
1 1 high 1 day results results-d row-ave ovmean within-v df

result
12

2 1 high 2 10 1 0 0 0 0 0 1
3 1 middle 1 8 2 0 0 0 0 1
4 1 middle 2 6
3 0 0 0 0 1
5 1 low 1 4
4 0 0 0 0 1
2
6 1 low 2 day
0
7 2 high 1 1 2 3 4
ANOVA SS df MSE F P Fcrit
8 2 high 2 between 0 3 0 #DIV/0! #DIV/0! 6.59
9 2 middle 1 Results Concentration Middle within 0 4 0
10 2 middle 2 12

11 2 low 1 10
total 0 7
12 2 low 2 8

result
6
13 3 high 1 4
14 3 high 2 2 Concentration middle
day
15 3 middle 1 0 day results results-d row-ave ovmean within-v df
16 3 middle 2 1 1.5 2 2.5 3 3.5 4
1 0 0 0 0 0 1
17 3 low 1 Results Concentration Low
2 0 0 0 0 1
18 3 low 2 12 3 0 0 0 0 1
19 4 high 1 10 4 0 0 0 0 1
20 4 high 2 8
result

21 4 middle 1 6
ANOVA SS df MSE F P Fcrit
22 4 middle 2 4
between 0 3 0 #DIV/0! #DIV/0! 6.59
2
23 4 low 1 day within 0 4 0
0
24 4 low 2 1 1.5 2 2.5 3 3.5 4

total 0 7

5 The repeatability and within laboratory reproducibility are calculated automatically for every
concentration level separately. Concentration low
day results results-d row-ave ovmean within-v df
Concentration repeatability within lab Comment 1 0 0 0 0 0 1
reproducibility 2 0 0 0 0 1
High 3 0 0 0 0 1
Middle 4 0 0 0 0 1
Low
ANOVA SS df MSE F P Fcrit
6 The repeatability and within laboratory reproducibility can be pooled over all concentration levels if the ratio between 0 3 0 #DIV/0! #DIV/0! 6.59
between the highest and lowest repeatability or within laboratory reproducibility is less than 3 (see the note below): within 0 4 0

Pooled repeatability (df=0)* total 0 7

Pooled within lab. reproducibility

Pooled repeatability/reproducibility
conc level r R r^2 R^2 df
high #VALUE! #VALUE!
middle #VALUE! #VALUE!
low #VALUE! #VALUE!
mean #VALUE! #VALUE!
sqrt(mean) #VALUE! #VALUE!

totdf 0

determining ratio for pooled r or R

ratio
rmax 0.00 #DIV/0!
rmin 0.00
Rmax 0.00 #DIV/0!
Rmin 0.00
3. Precision study to determine repeatability and within
laboratory reproducibility
Precision study in which the factors time & operator are changed
(for 2 analysts using 1 instrument)
Procedure
1 Take 3 samples with concentrations distributed over the working range.
2 Analyse these samples in duplicate on the same day.
3 Repeat these analyses on 4 different days giving in total 24 analytical results, and divide these
analysis over 2 operators (analysts) as listed in the set-up of the test.
4 Enter the results in the following table (blue fields). It is important that you fill in all fields.
Set-up of this test:

analysis nr day analyst concentration duplicate result Results Concentration High

Concentration high
1 1 1 high 1 day results results-d row-ave ovmean within-v df

result
12

2 1 1 high 2 10 1 0 0 0 0 0 1
3 1 1 middle 1 8
2 0 0 0 0 1
4 1 1 middle 2 6
3 0 0 0 0 1
4
5 1 1 low 1 4 0 0 0 0 1
2
6 1 1 low 2 0
day
7 2 2 high 1 1 2 3 4 ANOVA SS df MSE F P Fcrit
8 2 2 high 2 between 0 3 0 #DIV/0! #DIV/0! 6.59
9 2 2 middle 1 Results Concentration Middle within 0 4 0
10 2 2 middle 2 12

11 2 2 low 1 10 total 0 7
12 2 2 low 2 8

result
6
13 3 1 high 1
4
14 3 1 high 2 2
Concentration middle
day
15 3 1 middle 1 0 day results results-d row-ave ovmean within-v df
16 3 1 middle 2 1 1.5 2 2.5 3 3.5 4 1 0 0 0 0 0 1
17 3 1 low 1 2 0 0 0 0 1
Results Concentration Low
18 3 1 low 2 12
3 0 0 0 0 1
19 4 2 high 1 10
4 0 0 0 0 1
20 4 2 high 2 8

21 4 2 middle 1 ANOVA SS df MSE F P Fcrit

result

22 4 2 middle 2 4 between 0 3 0 #DIV/0! #DIV/0! 6.59

23 4 2 low 1 2
day
within 0 4 0
24 4 2 low 2 0
1 1.5 2 2.5 3 3.5 4
total 0 7

5 The repeatability and within laboratory reproducibility are calculated automatically for every
concentration level separately. Concentration low
day results results-d row-ave ovmean within-v df
Concentration repeatability within lab Comment 1 0 0 0 0 0 1
reproducibility 2 0 0 0 0 1
High 3 0 0 0 0 1
Middle 4 0 0 0 0 1
Low
ANOVA SS df MSE F P Fcrit
6 The repeatability and within laboratory reproducibility can be pooled over all concentration levels if the ratio between 0 3 0 #DIV/0! #DIV/0! 6.59
between the highest and lowest repeatability or within laboratory reproducibility is less than 3 (see the note below): within 0 4 0

Pooled repeatability (df=0)* total 0 7

Pooled within lab. reproducibility

Pooled repeatability/reproducibility
conc level r R r^2 R^2 df
high #VALUE! #VALUE!
middle #VALUE! #VALUE!
low #VALUE! #VALUE!
mean #VALUE! #VALUE!
sqrt(mean) #VALUE! #VALUE!

totdf 0

determining ratio for pooled r or R

ratio
rmax 0.00 #DIV/0!
rmin 0.00
Rmax 0.00 #DIV/0!
Rmin 0.00
4. Precision study to determine repeatability and within
laboratory reproducibility
Precision study in which the factors time & operator are changed
(for 2 analysts using their own instrument)
Procedure
1 Take 3 samples with concentrations distributed over the working range.
2 Analyse these samples in duplicate on the same day.
3 Repeat these analyses on 4 different days giving in total 24 analytical results, and divide these
analysis over 2 operators (analysts) as listed in the set-up of the test.
4 Enter the results in the following table (blue fields). It is important that you fill in all fields.
Set-up of this test:

analysis nr day analyst concentration duplicate result Results Concentration High

Concentration high
1 1 1 high 1 day results results-d row-ave ovmean within-v df

result
12

2 1 1 high 2 10 1 0 0 0 0 0 1
3 1 1 middle 1 8
2 0 0 0 0 1
4 1 1 middle 2 6
3 0 0 0 0 1
4
5 1 1 low 1 4 0 0 0 0 1
2
6 1 1 low 2 0
day
7 2 2 high 1 1 2 3 4 ANOVA SS df MSE F P Fcrit
8 2 2 high 2 between 0 3 0 #DIV/0! #DIV/0! 6.59
9 2 2 middle 1 Results Concentration Middle within 0 4 0
10 2 2 middle 2 12

11 2 2 low 1 10 total 0 7
12 2 2 low 2 8

result
6
13 3 1 high 1
4
14 3 1 high 2 2
Concentration middle
day
15 3 1 middle 1 0 day results results-d row-ave ovmean within-v df
16 3 1 middle 2 1 1.5 2 2.5 3 3.5 4 1 0 0 0 0 0 1
17 3 1 low 1 2 0 0 0 0 1
Results Concentration Low
18 3 1 low 2 12
3 0 0 0 0 1
19 4 2 high 1 10
4 0 0 0 0 1
20 4 2 high 2 8

21 4 2 middle 1 ANOVA SS df MSE F P Fcrit

result

22 4 2 middle 2 4 between 0 3 0 #DIV/0! #DIV/0! 6.59

23 4 2 low 1 2
day
within 0 4 0
24 4 2 low 2 0
1 1.5 2 2.5 3 3.5 4
total 0 7

5 The repeatability and within laboratory reproducibility are calculated automatically for every
concentration level separately. Concentration low
day results results-d row-ave ovmean within-v df
Concentration repeatability within lab Comment 1 0 0 0 0 0 1
reproducibility 2 0 0 0 0 1
High 3 0 0 0 0 1
Middle 4 0 0 0 0 1
Low
ANOVA SS df MSE F P Fcrit
6 The repeatability and within laboratory reproducibility can be pooled over all concentration levels if the ratio between 0 3 0 #DIV/0! #DIV/0! 6.59
between the highest and lowest repeatability or within laboratory reproducibility is less than 3 (see the note below): within 0 4 0

Pooled repeatability (df=0)* total 0 7

Pooled within lab. reproducibility

Pooled repeatability/reproducibility
conc level r R r^2 R^2 df
high #VALUE! #VALUE!
middle #VALUE! #VALUE!
low #VALUE! #VALUE!
mean #VALUE! #VALUE!
sqrt(mean) #VALUE! #VALUE!

totdf 0

determining ratio for pooled r or R

ratio
rmax 0.00 #DIV/0!
rmin 0.00
Rmax 0.00 #DIV/0!
Rmin 0.00
5. Precision study to determine repeatability and within
laboratory reproducibility
Precision study in which the factors time & instrument are changed
(for 2 analysts using 2 instruments)
Procedure
1 Take 3 samples with concentrations distributed over the working range.
2 Analyse these samples in duplicate on the same day.
3 Repeat these analyses on 4 different days giving in total 24 analytical results, and divide these
analysis over 2 instruments as listed in the set-up of the test.
4 Enter the results in the following table (blue fields). It is important that you fill in all fields.
Set-up of this test:

analysis nr day instrument concentration duplicate result Results Concentration High

Concentration high
1 1 1 high 1 day results results-d row-ave ovmean within-v df

result
12

2 1 1 high 2 10 1 0 0 0 0 0 1
3 1 1 middle 1 8
2 0 0 0 0 1
4 1 1 middle 2 6
3 0 0 0 0 1
4
5 1 1 low 1 4 0 0 0 0 1
2
6 1 1 low 2 0
day
7 2 2 high 1 1 2 3 4 ANOVA SS df MSE F P Fcrit
8 2 2 high 2 between 0 3 0 #DIV/0! #DIV/0! 6.59
9 2 2 middle 1 Results Concentration Middle within 0 4 0
10 2 2 middle 2 12

11 2 2 low 1 10 total 0 7
12 2 2 low 2 8

result
6
13 3 1 high 1
4
14 3 1 high 2 2
Concentration middle
day
15 3 1 middle 1 0 day results results-d row-ave ovmean within-v df
16 3 1 middle 2 1 1.5 2 2.5 3 3.5 4 1 0 0 0 0 0 1
17 3 1 low 1 2 0 0 0 0 1
12 Results Concentration Low
18 3 1 low 2 3 0 0 0 0 1
10
19 4 2 high 1 4 0 0 0 0 1
8
20 4 2 high 2
result

21 4 2 middle 1 6
ANOVA SS df MSE F P Fcrit
22 4 2 middle 2 4
between 0 3 0 #DIV/0! #DIV/0! 6.59
23 4 2 low 1 2
day within 0 4 0
24 4 2 low 2 0
1 1.5 2 2.5 3 3.5 4
total 0 7

5 The repeatability and within laboratory reproducibility are calculated automatically for every
concentration level separately. Concentration low
day results results-d row-ave ovmean within-v df
Concentration repeatability within lab Comment 1 0 0 0 0 0 1
reproducibility 2 0 0 0 0 1
High 3 0 0 0 0 1
Middle 4 0 0 0 0 1
Low
ANOVA SS df MSE F P Fcrit
6 The repeatability and within laboratory reproducibility can be pooled over all concentration levels if the ratio between 0 3 0 #DIV/0! #DIV/0! 6.59
between the highest and lowest repeatability or within laboratory reproducibility is less than 3 (see the note below): within 0 4 0

Pooled repeatability (df=0)* total 0 7

Pooled within lab. reproducibility

Pooled repeatability/reproducibility
conc level r R r^2 R^2 df
high #VALUE! #VALUE!
middle #VALUE! #VALUE!
low #VALUE! #VALUE!
mean #VALUE! #VALUE!
sqrt(mean) #VALUE! #VALUE!

totdf 0

determining ratio for pooled r or R

ratio
rmax 0.00 #DIV/0!
rmin 0.00
Rmax 0.00 #DIV/0!
Rmin 0.00
6. Precision study to determine repeatability and within
laboratory reproducibility
Precision study in which the factors time, operator & equipment are changed
(for more than 2 combinations of analysts & instruments)
Procedure
1 Take 3 samples with concentrations distributed over the working range
2 Choose 4 different combinations of analysts and instruments out of all possible analyst/instrument combinations.
The 4 combinations must contain as much different analysts and instruments as possible (see example).

Example: Total solids analysis is performed on laboratory X by 4 analysts that use 3 total solids analysers

Possible analyst/instrument combinations:

Analyst Instrument Combinations The 4 combination marked with a + sign is an example of

1 1 % 4 combinations that are suitable. These combinations
1 2 include all possible analysts and instruments
1 3 +
2 1 % If for example the 4 combinations marked with a % sign
2 2 % are chosen then this is not OK. These combinations do
2 3 + not include Analyst 4 and Instrument 3.
3 1 %
3 2 + If more than 4 analysts and/or instruments are present
3 3 the "redundant" analysts and/or instruments are not used.
4 1 + However if a deviation from the official method is obtained
4 2 one can repeat the precision study with the analysts and/or
4 3 instruments not included in the current test to investigate
the cause of this deviation

3 Each of the 4 analyst/instrument combinations analyses the 3 samples in duplicate on one day
giving in total 24 analytical results.
4 Enter the analyst/instrument combinations and the results in the following table (blue fields).
It is important that you fill in all result fields.
Set-up of this test:

analysis nr day analyst equipment concentration duplicate result Concentration high

Results Concentration High
1 1 high 1 day results results-d row-ave ovmean within-v df

result
12

2 1 high 2 10 1 0 0 0 0 0 1
3 1 middle 1 8 2 0 0 0 0 1
4 1 middle 2 6 3 0 0 0 0 1
5 1 low 1 4
4 0 0 0 0 1
2
6 1 low 2 day
0
7 2 high 1 1 2 3 4
ANOVA SS df MSE F P Fcrit
8 2 high 2 between 0 3 0 #DIV/0! #DIV/0! 6.59
9 2 middle 1 Results Concentration Middle within 0 4 0
10 2 middle 2 12

11 2 low 1 10 total 0 7
12 2 low 2 8
result

6
13 3 high 1 4
14 3 high 2 2 Concentration middle
day
15 3 middle 1 0 day results results-d row-ave ovmean within-v df
16 3 middle 2 1 1.5 2 2.5 3 3.5 4
1 0 0 0 0 0 1
17 3 low 1 Results Concentration Low
2 0 0 0 0 1
18 3 low 2 12 3 0 0 0 0 1
19 4 high 1 10 4 0 0 0 0 1
20 4 high 2 8

21 4 middle 1 ANOVA SS df MSE F P Fcrit

result

22 4 middle 2 4
between 0 3 0 #DIV/0! #DIV/0! 6.59
2
23 4 low 1 day within 0 4 0
0
24 4 low 2 1 1.5 2 2.5 3 3.5 4

total 0 7

5 The repeatability and within laboratory reproducibility are calculated automatically for every
concentration level separately. Concentration low
day results results-d row-ave ovmean within-v df
Concentration repeatability within lab Comment 1 0 0 0 0 0 1
reproducibility 2 0 0 0 0 1
High 3 0 0 0 0 1
Middle 4 0 0 0 0 1
Low
ANOVA SS df MSE F P Fcrit
6 The repeatability and within laboratory reproducibility can be pooled over all concentration levels if the ratio between 0 3 0 #DIV/0! #DIV/0! 6.59
between the highest and lowest repeatability or within laboratory reproducibility is less than 3 (see the note below): within 0 4 0

Pooled repeatability (df=0)* total 0 7

Pooled within lab. reproducibility

Pooled repeatability/reproducibility
conc level r R r^2 R^2 df
high #VALUE! #VALUE!
middle #VALUE! #VALUE!
low #VALUE! #VALUE!
mean #VALUE! #VALUE!
sqrt(mean) #VALUE! #VALUE!

totdf 0

determining ratio for pooled r or R

ratio
rmax 0.00 #DIV/0!
rmin 0.00
Rmax 0.00 #DIV/0!
Rmin 0.00
LIMITS OF DETECTION AND QUANTIFICATION
Towards the lower end of the range of the analytical method it becomes increasingly difficult to
distinguish the response of the test sample from the response of a blank, or in the case of
two-dimensional patterns (spectra, chromatograms) from the base line. The larger the difference
between the 2 responses, the more confident we are that the response of the test sample is due
to the genuine presence of the sought component and not an artefact of blank or baseline
variations (noise). The observed difference is balanced against the random variations in the
blank to arrive at a statistically significant positive conclusion.
Limits of detection and quantitation are important criteria in trace analysis.

Important definitions for this study:

The capability of detection, indicates the lowest amount which is detectable by all means.
This value should be reported when no significant signal has been detected. Therefore we need
a criterion to decide whether a signal has been detected or not: the decision limit. The decision
limit is expressed in signal units (e.g. peak area, mV,..).
The detection limit, is the amount or concentration which corresponds to the decision limit.
The limit of quantitation indicates the lowest amount which can be determined with a relative
standard deviation of 10%.

All three limits are based on an estimate of the standard deviation at very low (signal) levels or of a blank sample.
A reliable estimate of this standard deviation is very essential and critical as it highly influences the calculated
limits of detection. For instance, its value may depend on whether replicates have been measured under repeatability
or within-laboratory reproducibility conditions.
One should also carefully consider the definition of the "signal". For instance when in chromatography the signal has
been defined as a peak area, it is not allowed to consider the baseline noise as a blank signal. In some other cases
the net signal is calculated as the difference between 2 signals, where one signal is usually a blank signal.

The use of blank samples (contains all ingredients except the analyte) is highly preferred over the use of a blank
solution (solvent) in determining the limit of detection and quantification. In order to avoid any ambiguity it is
highly recommended to analyse test portions of a fortified sample blank, with the lowest concentration
giving a detectable signal in all cases.

Procedure:
1 Take a test sample that contains the sought component in very low concentration, slightly above the
expected capability of detection.
If such a test sample is not available, make a suitable test sample by adding the sought component to a
blank sample in the desired concentration.
2 Divide the test sample in 10 roughly equal test portions.
3 Analyse each test portion once according to the analytical method that is to be validated. All test portions
should undergo the complete analytical procedure, including sample preparation.
4 Enter the results in the table below (blue fields).

test analytical
portion result
1
2
3
4
5
6
7
8
9
10
average
stand.dev

5 Calculations

Detection limit
Capability of Detection
Limit of Quantitation

If the Capability of Detection is more than 3 standard deviations away from the average, the test
should be repeated with a sample closer to the Capability of Detection.

RESULT TEST
Conclusion
| Average - Capability of Detection | =
stand.dev

6 Further calculations:
Possibly a decision limit can be calculated from the detection limit by using the standard algorithm
used to calculate the results from the signal (this is strongly dependent from the analytical procedure
and local circumstances)

7 Conclusions and report; what to do with these results.

If the result of the analysis of a sample is lower than the detection limit (result<detection limit, thus
signal<decision limit) report result as: Concentration < Capability of detection.

If the result of the analysis of a sample is between the detection limit and the capability of detection
(detection limit<result<capability of detection) report result with a 95% confidence interval, calculated
as follows:
CI= x +/- 2.262*s/√n
x=the analytical result
s=standard deviation of the 10 analytical results
n=number of observations used to obtain the analytical result

If the result of the analysis of a sample is above the capability of detection (result>capability of detection)
report result with a confidence interval calculated using the appropriate repeatability defined
in a precision study.
The confidence interval is now calculated as follows:
CI= x +/- 2.179*r/(2.8*√n )
x=the analytical result
r=repeatability
n=number of observations used to obtain the analytical result

Enter the requested information below (blue fields) to obtain an advice for reporting
(the table in procedure section 4 above should be filled in as well):

result x
n
repeatability r

What to report in your case:

LINEAR WORKING RANGE
For any quantitative method it is necessary to determine the range of analyte concentrations over which
the method may be applied. At the lower end of the concentration range the limiting factor is the value
of the limit of quantification. At the upper end of the concentration range various limitations may be imposed
depending on the detection mechanism.

Within the working range there may exist a linear range, where the detector response will have a linear
relationship to the analyte concentration. The extend of this range should be established during the
evaluation of the working range.

To remain practical it is recommended to check the linearity within the

concentration range where the samples are expected.
Evaluation of the linear range is also useful for designing a calibration strategy when using the method
on a day-to-day basis. Within the linear range, one calibration point may be sufficient, provided that
the slope of the calibration line is known.
The working range and linear range can differ for different sample types.
What has to be analysed depends very much on the information already available from the other
validation steps. For instance during the trueness study, spiked samples at various levels may already
be measured.

Procedure
1 Take at least 8 test samples with concentrations distributed over the concentration range where the samples
are expected. It is also possible to spike blank samples with different known amount of the component
to be analysed to obtain the samples.
2 Analyse these samples in duplicate with the method that is to be validated.
Each test portion must undergo the complete analytical procedure including sample preparation.
The samples must be analysed in random order
3 Enter the results in the following table (blue fields). The fields that are not used must be left empty

Analysis: Least Squares Regression

test Concentration results (duplicates)
f(x) = NaN
12 x + NaN
1 2 R² = 0 x y1 y2 yaverow xi-xave (xi-xave)2 yi-yave (yi-yave)2 (xi-xave)(yi-yave) ycal resy1 resy2 e2i=(resy1+resy2)2/2 d2i=(y1-y2)2/2
1
10
2
3
4 8
5
results

6 6
7
8 4
9
10 2
11
12 0
13 0 2 4 6 8 10 12
14
15 concentration
16
total 0.00 0.00 0.00 0.00 0.00 0.00 0 0 0
average #DIV/0! #DIV/0! SSLOF SSPE
4 The linearity of the results is automatically checked with a Lack-of-Fit test (the correlation coefficient nr of levels 0
which is normally used is a poor measure of linearity). r: #DIV/0!
b:(slope) #DIV/0!
RESULT LACK-OF-FIT TEST a:(intercept) #DIV/0! F= 0
Conclusion
F 0.00 Err:502

Fcritical Err:502 please note that this is only valid if you meet the requirements under point 5
This table is derived from: Statistics for Analytical Chemistry
5 It should be checked if the standard deviation of this analysis is not deviating significantly from the results of the precision study 3rd edition
This can be checked by the following test: J.C. Miller and J.N. Miller
Fill in the next fields the pooled repeatability and degrees of freedom (df) to match found in the precision study

Pooled repeatability
df Fcritical for s test Err:502
result test #DIV/0!
If this test shows that "LOF is valid" continue with point 6. If this test shows that "LOF is NOT valid" investigate
the difference in standard deviation between this analysis and the precision study before continuing

6 If the lack-of-fit test shows that "the method is linear over the investigated range" the test can be stopped. However
when the range was not chosen well, i.e. the investigated range was not wider then the range where samples may
be expected under normal conditions, the range should be extended. In that case add a point at the upper or
lower end of the inspected range and repeat procedure step 2 (the duplicate analysis). Next repeat the lack-of fit
test.
In certain situations it is necessary to define the boundaries of the range in between its behaviour is still linear. In
those situations new samples should be added at either the upper or lower end of the range untill the lack-of-fit
test fails.

If the lack-of-fit test shows that "the method is NOT linear over the investigated range", remove a point at the
upper or lower end of the inspected range (see graph!). Continue the removal of a point until the Lack-of-fit
test indicates that "the method is linear over the investigated range". If 2 or more observations were deleted
it is advised to analyse additional samples in the reduced concentration range.
Validation of this software package

This program was developed by Quality Systems, the research programme of the Foods
Quality Assurance Group (FQAG) in 2002.

For questions about this program please contact:

P.A. Zandbelt
Unilever Research & Development Vlaardingen
Postbus 114
3130AC Vlaardingen
tel: 0031-(0)10-4606205
fax: 0031-(0)10-4605310
E-mail: Peter.Zandbelt@Unilever.com

Purpose of this program

The purpose of this program is to help users with the statistical calculations of an
in-house validation. The program gives some information about the validation
but it is by no means a validation tool. This means that interpretation of the data
and the set-up of the validation strategy remains the analysts responsibility.
Validation in the context of this program means ensuring that the method is
implemented in the laboratory performs as it was designed (e.g by official
organisations such as ISO and AOCS). The program only allows a fixed
set-up of validation. It does by no means replace the official method validation
standard (ref....) and should be used in addition to it

Responsibility
Microsoft Excel was used for pragmatic reasons: it is easy to use and available on all
Unilever computers. Although a lot of effort has been put in testing to ensure that it
works properly, the developers of this package can not be held responsible for potential
malfunctioning and related consequences.

Testing
To provide some assurance for proper functioning of the calculations performed in this
program, several data sets are provided for testing. Testing must be performed at least
once, after installation of this program on your computer and prior to using the program
on real data.

Testdatasets

The following test datasets must be used in sheet:

1. Repeatability study

TEST DATASET 1 TEST DATASET 2 TEST DATASET 1

concentration result concentration result concentration result
level level level
1 1.13 1 1 1 2.24
1.23 1 2.7
1.22 1
1.2 1
1.12 1
2 6.55 2 2 2 7.56
7.98 2 6.99
6.54 2
6.37 2
7.96 2
3 29.7 3 3 3 30.8
30 3 29.5
30.1 3
29.5 3
29.1 3
4 211 4 4 4
204 4
212 4
213 4
205 4

OUTPUT TEST DATASET 1 OUTPUT TEST DATASET 2 OUTPUT TEST DATASET 3

standard repeatability (r) overall standard repeatability (r) overall standard repeatability (r) overall
deviation per conc. level repeatability deviation per conc. level repeatability deviation per conc. level repeatability
1 0.051 0.144 5.994 1 0.000 0.000 0.000 1 0.325 0.911 1.706
2 0.816 2.284 2 0.000 0.000 2 0.403 1.129
3 0.402 1.127 3 0.000 0.000 3 0.919 2.574
4 4.183 11.713 4 0.000 0.000 4

The following test datasets must be used in sheets:

2. Precision study to determine repeatability and within laboratory reproducibility
Precision study in which only the factor time is changed (for 1 analyst using 1 instrument)
3. Precision study to determine repeatability and within laboratory reproducibility
Precision study in which the factors time & operator are changed (for 2 analysts using 1 instrument)
4. Precision study to determine repeatability and within laboratory reproducibility
Precision study in which the factors time & operator are changed (for 2 analysts using their own instrument)
5. Precision study to determine repeatability and within laboratory reproducibility
Precision study in which the factors time & instrument are changed (for 1 analyst using 2 instruments)
6. Precision study to determine repeatability and within laboratory reproducibility
Precision study in which the factors time, operator & equipment are changed (for 2 or more analysts & 2 or more instruments)

TEST DATASET 1 TEST DATASET 2 TEST DATASET 3

analysis nr result analysis nr result analysis nr result
1 12 1 21.48 1 25
2 12.1 2 19.18 2 25
3 6 3 9.69 3 10
4 6.1 4 12.22 4 10
5 3 5 6.18 5 5
6 3 6 5.69 6 5
7 13 7 19.35 7 25
8 12.5 8 21.5 8 25
9 5.9 9 9.55 9 10
10 5.9 10 8.61 10 10
11 2.8 11 6.16 11 5
12 2.9 12 3.39 12 5
13 12.2 13 22.9 13 25
 14 12.1 14 20.03 14 25
15 6.5 15 7.22 15 10
16 6.5 16 7.54 16 10
17 3 17 3.85 17 5
18 2.9 18 6.19 18 5
19 13 19 21.5 19 25
20 13.2 20 21.24 20 25
21 6 21 8.1 21 10
22 6.5 22 9.18 22 10
23 3 23 5.1 23 5
24 3.2 24 3.95 24 5

OUTPUT TEST DATASET 1 OUTPUT TEST DATASET 2 OUTPUT TEST DATASET 3

Concentration repeatability within lab Concentration repeatability within lab Concentration repeatability within lab
reproducibility reproducibility reproducibility
High 0.55 1.45 High 3.56 3.56 * High 0.00 0.00
Middle 0.50 0.81 Middle 2.90 4.63 Middle 0.00 0.00
Low 0.24 0.34 Low 3.28 3.28 * Low 0.00 0.00

Comment: Comment: Comment:

None *This is the best estimate of the within lab. reproducibility, none
but most probably this value is higher. This error is
caused by the variance between duplicates being
larger than between days.

pooled repeatability 0.45 pooled repeatability 3.26 pooled repeatability 0

pooled reproducibility 0.98 pooled reproducibility 3.87 pooled reproducibility 0
df=12 df=18 df=21

NOTES: NOTES NOTES

pooling is statistically justified for repeatability pooling is statistically justified for repeatability #DIV/0!
pooling is statistically NOT justified for laboratory reproducpooling is statistically justified for laboratory reproducibili #DIV/0!

The following test datasets must be used in sheet:

Limits of detection and quantification

TEST DATASET 1 TEST DATASET 2

test portion result test portion result
1 0.035 1 1
2 0.023 2 1.1
3 0.017 3 1
4 0.045 4 0.9
5 0.033 5 0.8
6 0.02 6 1.2
7 0.041 7 1
8 0.026 8 1.05
9 0.013 9 1.1
10 0.022 10 0.9

OUTPUT TEST DATASET 1 OUTPUT TEST DATASET 3

Limit of Detection 0.032 Limit of Detection 0.350
Capability of Detection 0.063 Capability of Detection 0.699
Limit of Quantification 0.106 Limit of Quantification 1.165
Result test 3.40 Result test 2.62
Comment: Comment:
It is advised to repeat the test with a sample closer to The average is acceptably close to
0.063 0.699

The following test datasets must be used in sheet:

Linear working range

TEST DATASET 1 TEST DATASET 2 TEST DATASET 3

Concentration results 1 result 2 Concentration results 1 result 2 Concentration results 1 result 2
4 0.26 0.28 4 0.26 0.28 4 0.26 0.28
4.5 0.36 0.37 4.5 0.36 0.37 4.5 0.36 0.37
5 0.49 0.48 5 0.49 0.48 5 0.49 0.48
5.5 0.6 0.62 5.5 0.6 0.62 5.5 0.6 0.62
6 0.7 0.69 6 0.7 0.69 6 0.7 0.69
6.5 0.83 0.82 6.5 0.83 0.82 6.5 0.83 0.82
7 0.9 0.91 7 0.9 0.91
7.5 1.05 1.1 7.5 1.05 1.1
8 1.15 1.16 8 1.15 1.16
8.5 1.24 1.26 8.5 1.24 1.26
9 1.36 1.37 9 1.36 1.37
9.5 1.47 1.48 9.5 1.47 1.48
10 1.6 1.58 10 1.6 1.58
10.5 1.68 1.7 10.5 1.68 1.7
11 1.79 1.81 11 1.79 1.81
11.5 1.9 1.9 11.5 1.85 1.83

OUTPUT TEST DATASET 1 OUTPUT TEST DATASET 2 OUTPUT TEST DATASET 3

RESULT LACK-OF-FIT TEST RESULT LACK-OF-FIT TEST RESULT LACK-OF-FIT TEST
F 2.55 F 2.30 F 5.33
Fcritical 4.53 Fcritical 2.37 Fcritical 2.37
Conclusion The method is linear over the Conclusion The method is linear over the Conclusion The method is NOT linear over
investigated range investigated range the investigated range

The following test datasets must be used in sheet:

1. Trueness study with a certified reference material

TEST DATASET 1 TEST DATASET 2 TEST DATASET 3

Result Concentration Result Concentration Result Concentration
of CRM of CRM of CRM
990 1000 1000 1000 802 1000
900 1000 805
995 1000 812
997 1000 799
1005 1000 803
985 1000 800
OUTPUT TEST DATASET 1 OUTPUT TEST DATASET 2 OUTPUT TEST DATASET 3
average 978.666666667 average 1000 average 803.5
standdev 39.1237353363 standdev 0 standdev 4.679743582719
test value 1.33565419381 test value test value 102.852800789
table value (6) 2.571 table value (6) 2.571 table value (6) 2.571
Conclusion no statistical Conclusion no statistical Conclusion bias with 95%
significant bias significant bias confidence

The following test datasets must be used in sheets:

2. Trueness study with a blank sample spiked with standards or CRM
3. Trueness study with a test sample spiked with standards or CRM

TEST DATASET 1 TEST DATASET 2

known amount result 1 result 2 known amount result 1 result 2
of added of added
component component
blank 0 0.01 0.02 blank 0 5.3 5.2
low 1 1.1 1.2 low 1 6.3 6.5
intermediate 10 9.7 12.1 intermediate 10 17.7 18.5
high 100 98.5 101 high 100 110.5 112

OUTPUT TEST DATASET 1 (for 2. Trueness with a blank sample.....) OUTPUT TEST DATASET 2 (for 2. Trueness with a blank sample.....)
intercept 0.3558 intercept 6.0024
95% Confidence interval intercept/2 1.1246 95% Confidence interval intercept/2 1.2251
slope 0.9945 slope 1.0540
95% Confidence interval slope/2 0.0224 95% Confidence interval slope/2 0.0244
GENERAL CONCLUSION no bias detected GENERAL CONCLUSION both a proportional and off-set
(proportional as well as off-set) bias are detected
The maximum off-set bias 1.4804 The maximum off-set bias % 7.2274
The maximum proportional bias 0.0279 The maximum proportional bias % 0.0783

OUTPUT TEST DATASET 1 (for 3. Trueness with a test sample.....) OUTPUT TEST DATASET 2 (for 3. Trueness with a test sample.....)
intercept 0.3558 intercept 6.0024
slope 0.9945 slope 1.0540
95% Confidence interval slope/2 0.0224 95% Confidence interval slope/2 0.0244
GENERAL CONCLUSION no proportional bias detected GENERAL CONCLUSION a proportional bias is detected
The maximum proportional bias % 2.7858% The maximum proportional bias % 7.8347%
Recovery at low concentration (%) 113.50% Recovery at low concentration (%) 115.00%
Recovery at intermediate concentration (%) 108.85% Recovery at intermediate concentration (%) 128.50%
Recovery at high concentration (%) 99.74% Recovery at high concentration (%) 106.00%

The following test datasets must be used in sheet:

4. Trueness study with a CRM spiked with standards

TEST DATASET 1 TEST DATASET 2

known amount result 1 result 2 known amount result 1 result 2
of added of added
component component
blank 5 0.01 0.02 blank 5 5.3 5.2
low 6 1.1 1.2 low 6 6.3 6.5
intermediate 15 9.7 12.1 intermediate 15 17.7 18.5
high 105 98.5 101 high 105 110.5 112

OUTPUT TEST DATASET 1 (for 4. Trueness with a CRM.....) OUTPUT TEST DATASET 2 (for 4. Trueness with a CRM.....)
intercept -4.6168 intercept 0.7325
95% Confidence interval intercept/2 1.1900 95% Confidence interval intercept/2 1.2964
slope 0.9945 slope 1.0540
95% Confidence interval slope/2 0.0224 95% Confidence interval slope/2 0.0244
GENERAL CONCLUSION an off-set bias is detected GENERAL CONCLUSION a proportional bias is detected
(no proportional bias) (no off-set bias)
The maximum off-set bias 5.8068 The maximum off-set bias 2.0289
The maximum proportional bias 0.0279 The maximum proportional bias 0.0783
Recovery at low concentration (%) 113.50% Recovery at low concentration (%) 115.00%
Recovery at intermediate concentration (%) 108.85% Recovery at intermediate concentration (%) 128.50%
Recovery at high concentration (%) 99.74% Recovery at high concentration (%) 106.00%

The following test datasets must be used in sheet:

5. Trueness study by comparison with a reference method on one concentration level

TEST DATASET 1 TEST DATASET 2 TEST DATASET 3

Test Result with Result with Test Result with Result with Test Result with Result with
portion unvalidated reference portion unvalidated reference portion unvalidated reference
method method method method method method
1 3.9 5.2 1 5 5.2 1 3 3
2 4.9 5.1 2 4.9 5.1 2 3 3
3 3.9 5 3 4.8 5 3 3 3
4 4.9 5.2 4 5.1 5.2 4 3 3
5 3.9 4.9 5 5.2 4.9 5 3 3
6 4.9 5 6 5 5 6 3 3

OUTPUT TEST DATASET 1 OUTPUT TEST DATASET 2 OUTPUT TEST DATASET 3

unvalidated average 4.4000 unvalidated average 5.0000 unvalidated average 3.0000
method standdev 0.5477 method standdev 0.1414 method standdev 0.0000
reference average 5.0667 reference average 5.0667 reference average 3.0000
method standdev 0.1211 method standdev 0.1211 method standdev 0.0000
bias 0.6667 bias 0.0667 bias 0.0000
standdev bias 0.2290 standdev bias 0.0760 standdev bias 0.0000
critical distanc 0.5102 critical distanc 0.1694 critical distanc 0.0000
conclusion: bias with 95% confidence conclusion: no statistical significant bias conclusion: no statistical significant bias
Biasmax 0.1769 Biasmax 0.2360 Biasmax 0.0000

The following test datasets must be used in sheet:

6. Trueness study by comparison with a reference method at various concentration levels

TEST DATASET 1 TEST DATASET 2 TEST DATASET 3

Test Result with Result with Test Result with Result with Test Result with Result with
portion reference unvalidated portion reference unvalidated portion reference unvalidated
method method method method method method
1 1 1 1 0.88 1 1 0.98 1
 2 2.2 2 2 1.78 2 2 2.08 2
3 3 3 3 2.75 3 3 3.12 3
4 3.9 4 4 3.85 4 4 4.17 4
5 4.8 5 5 4.76 5 5 5.25 5
6 6.2 6 6 5.65 6 6 6.33 6

method standard method standard method standard

deviation deviation deviation
to be validated 1.4 to be validated 1.4 to be validated 1.4
reference 2.3 reference 2.3 reference 2.3

OUTPUT TEST DATASET 1 OUTPUT TEST DATASET 2 OUTPUT TEST DATASET 3

average standard deviation average standard deviation average standard deviation
reference 3.52 1.8616 reference 3.28 1.8124 reference 3.66 1.9944
unvalidated 3.5 1.8708 unvalidated 3.5 1.8708 unvalidated 3.5 1.8708

dave -0.0167 dave 0.22167 dave -0.155

sd 0.1602 sd 0.0813 sd 0.1244
CD 0.16809 CD 0.08534 CD 0.13055
λ 0.3705 λ 0.3705 λ 0.3705
r 0.99633 r 0.99953 r 0.99997
slope+CIslope/2 1.01+0.11 slope+CIslope/2 1.03+0.04 slope+CIslope/2 0.94+ 0.01
intercept -0.0401 intercept 0.11511 intercept 0.07135

Conclusion: no bias detected (proportional Conclusion: an off-set bias is detected (conc Conclusion: a proportional bias is detected
as well as off-set) independent) (off-set cannot be tested)
Biasoffset -0.0167+/-0.1682 Biasoffset 0.2217+/-0.0854 Biasoffset -0.1550+/-0.1305
Max(Biasoffset) 0.1848 Max(Biasoffset) 0.3070 Max(Biasoffset) 0.2855
Biasprop 1.0067+/-0.1056 Biasprop 1.0325+/-0.0373 Biasprop 0.9381+/-0.0095
Max(Biasprop) (%) 11.2293 Max(Biasprop) (%) 6.9807 Max(Biasprop) (%) 7.1442