INTERNATIONAL ISO

STANDARD 8466-1
First edition
1990-03-01

Water quality - Calibration and evaluation of

analytical methods and estimation of
Performance characteristics

Part 1:
Statistical evaluation of the linear calibration function

Qualith de l’ea r- Etalonnage et Evaluation des mhthodes d’analyse et estimation

des caractkes de Performance
Partie 1: haha tion s ta tistique de Ia fonction lin&aire d’h talonnage

Reference number
Licensed to LAAM /MANUEL SOARES ISO 8466-1 : 1990 (EI
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
ISO 8466-1 : 1990 (EI

Contents Page
.. .
Foreword ............................................................ Ill

1 Scope .............................................................. 1

2 Definitions.. ........................................................ 1

3 Symbols.........................................................~ .. 2

4 Performance ........................................................ 3

4.1 Choice of working range .......................................... 3

4.2 Calibration and characteristics of the method ........................ 4

4.3 Assessment .................................................... 5

5 Example ............................................................ 6

5.1 Choice of working range .......................................... 6

5.2 Calibration and characteristics of the method ........................ 7

5.3 Evaluation ...................................................... 7

Annex A Bibliography ................................................. 8

0 ISO 1990
All rights reserved. No patt of this publication may be reproduced or utilized in any form or by any
means, electronie or mechanical, including photocopying and microfilm, without Permission in
writing from the publisher.
International Organization for Standardization
Case postale 56 l CH-1211 Geneve 20 l Switzerland
Printed in Switzerland Licensed to LAAM /MANUEL SOARES
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
ii
 ISO 8466-1 : 1990 (E)

Foreword
ISO (the International Organization for Standardization) is a worldwide federation of
national Standards bodies (ISO member bedies). The work of preparing International
Standards is normally carried out through ISO technical committees. Esch member
body interested in a subject for which a technical committee has been established has
the right to be represented on that committee. International organizations, govern-
mental and non-governmental, in liaison with ISO, also take patt in the work. ISO
collaborates closely with the International Electrotechnical Commission (IEC) on all
matters of electrotechnical standardization.

Draft International Standards adopted by the technical committees are circulated to

the member bodies for approval before their acceptance as International Standards by
the ISO Council. They are approved in accordance with ISO procedures requiring at
least 75 % approval by the member bodies voting.

International Standard ISO 8466-1 was prepared by Technical Committee ISO/TC 147,
Water quaiity.

ISO 8466 consists of the following Parts, under the general title Water gua/ity -
Calibration and evaluation of analytical methods and estimation of Performance
charac teris tics :

- Part 7: Statisticaf evaluation of the linear calibration function

- Part 2: Calibration strategy for non-linear calibration functions

- Part 3: Method of Standard addition

- Part 4: Estimation of Limit of detection and limit of determination of an analytical

basis method.

Annex A of this part ISO 8466 is for information only.

Licensed to LAAM /MANUEL SOARES

ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
 This page intentionally left blank

Licensed to LAAM /MANUEL SOARES

ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
INTERNATIONAL STANDARD ISO 8466-1 : 1990 (E)

Water quality - Calibration and evaluation of analytical

methods and estimation of Performance characteristics

Part 1:
Statistical evaluation of the linear calibration function

1 Scope
1 Originai Sample 1
This part of ISO 8466 describes the Steps to be taken in
evaluating the statistical characteristics of the linear calibration
_ , MeasUri!g ;X~~nstructi;
function. lt is applicable to methods requiring a calibration. Fur-
ther Parts of ttis lnternationa~ Standard will cover the deter-
mina$.i!! ok IGr%t of detection and FEmit of determination, the
effect of Ln$erferences and other performante chara,cteristics

lt is intended especially for the evaluation of the pure analytical

1
method and for the calculation of Performance characteristics measuring instructions
of the calibration function.

In Order to derive comparable analytical results and as a basis 1 Measur; value 1 ’

for analytical quality control the calibration and evaluation of
analytical methods have to be performed uniformly.

r-YG&q
calibrating and evaluating instructions

2 Definitions
For the purposes of this part of ISO 8466, the following defini-
tions apply.

2.1 analytical method: An analytical method is composed

of procedural, measuring, calibrating and evaluating instruc- Figure 1 - The analytical method
tions (see figure 1).
23 evaluating instruction : A calculation guide for the
Whereas the procedural and measuring instructions depend on computation of concentrations from the measured values by
the method, and are therefore the Object of standardization of the use of the calibration function. Additionally, the confidence
the respective method, the calibrating and evaluating instruc- range permits an objective assessment of the imprecision of the
tions are valid for any analytical method requiring calibration. analytical result[*J.

2.2 calibrating instruction : Describes the approach to 2.4 measured values : The concentration-dependent initial
determine the calibration function from information values, yi, values (e.g. extinction) of a measuring System.
obtained by measuring given Standard concentrations, xi. The
slope of the calibration function, b, as a measure of sensitivity NOTE - Information value and measured volume are synonymous.
of the analytical method and the Standard deviation of the
method, sXO, are figures of merit and characteristics which 25 residual Standard deviation, s : The residual Standard
result from the calibration experiment. deviation describes the scatter of the information values about
the calculated regression line. lt is a figure of merit, describing
The Standard deviation, sXO, allows the comparison of indepen- the precision of the calibration.
dent analytical methods.
For the purpose of this Standard, the Standard deviation of the
For the user of the method, these characteristics present method means the Standard of deviation of the calibration pro-
criteria for the internal laboratory quality control. cedure.
Licensed to LAAM /MANUEL SOARES
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
1
 ISO 8466-1 : 1990 (EI

2.6 Standard deviation of the method sXO: The ratio of Yi Information value of the Standard concentra-
the residual Standard deviation, s,,, to the sensitivity of the tion Xi calculated from the calibration function.
calibration function, b. lt is a figure of merit for the perfor-
mance of the analytical method, and is valid within the working s2
i
Variante of the information values for the
range (sec equation 13). analyses of Standard samples, having the con-
centration Xi.
For the purpose of this Standard, the Standard deviation of the
method means the Standard of deviation of the calibration pro-
fi Degrees of freedom for the calculation of the
cedure. variance (f i = tli- 1).

2.7 coefficient of Variation of the method, VXO: The ratio a Calculated blank (Ordinate intercept of the cali-
of the Standard deviation of the method sXOto the appertaining bration straight line).
mean, F, which is the centre of the working range.

See also note to 2.5 and 2.6. b Sensitivity of the method (slope of the cali-
bration line; coefficient of regression
2.8 working range (of an analytical method): The interval,
being experimentally established and statistically proved by the x Mean of the Standard concentrations
resulting from the calibration experiment.
calibration of the method, between the lowest and highest
quantity or mass concentration. The lowest possible limit of a
working range is the limit of detection of an analytical method. Y Mean of the information values Yit resulting
from the calibration experiment.
2.9 homogeneity of variances: Homogeneity of variances
Residual Standard deviation.
of pooled data, such as those resulting from replicate analyses sY
at different levels, is confirmed if these variances are not
Residual Standard deviation obtained by linear
significantly correlated to their appertaining concentrations. sY1
regression calcula tion.
2.10 sensitivity of the analytical method: The slope of
sY2 Residual Standard deviation obtained by non-
the calibration function of the complete analytical method, in-
linear regression calculation.
clusive of all procedural Steps, within the working range in
question.
DS2 Differente of variances.

2.11 measuring Sample (reaction Sample) : A Sample which

Y Information value of an analysed Sample.
tan be directly submitted to the measurement of the determi-
nand. A measuring Sample is normally obtained by adding the
n Number of replicates on the same analysed
required reagents to the analytical Sample. Obviously, measur-
Sample.
ing and analytical Sample are identical if no reagents have to be
added to the analytical Sample.
P Mean of information values, resulting from n
replicates.

3 Symbols x Concentration sf the analytical Sample,

Concentration of the Ph Standard Sample. calculated from the i nformation value Ym
Xi

i Subscript of the concentration levels, where 2 Concentration of the analytical Sample,

i= 1,2, . . . . N. calculated from the mean of the information
h

values 7.
N Number of concentration levels (for this part of
ISO 8466, N = IO). tV;, l-a) Tabled value of the t-distribution with fl =
N - 2 degrees of freedom and a confidence
Xl Concen tration of the Standard Sample at the lower level of (1 - a) (f-factor of Student’s distribu-
level of the working range ( 1st Standard Sample). tion).

Concentration of the Standard Sample at the

XI0
upper level of the working range (10th Standard
FV;,f2,1 -a) Tabled value of the F-distribution (Fisher-
Snedecor) with fl and f2 degrees of freedom
Sample).
and a confidence level of (1 - a).

Yi,j jth information value for the concentration Xi.

%O Standard deviation of the method.
j Subscript of the replicate j of level i, where j = 1,
2, . . . . ni. VXO Coefficient of Variation of the method.

% Number of replicates per level Xi. VB (x, Confidence interval for the concentration .?.

Mean of the information values Yi,j of Standard Confidence interval of the mean Z?of the con-
samples, having the concentration Xi. centration.
Licensed to LAAM /MANUEL SOARES
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
2
 ISO 8466-1 :, 1990 IE)

4 Performance 4.1.2 Test for homogeneity of the variances

4.1 Choice of working range Both data sets of the concentrations x1 and xlo are used t~
calculate the variances sf and s$, as given in equation (1) :
Esch calibration experiment is started with the choice of a 10
preliminary working rangeL3].
t: (Y i t j - El*
j=l
The working range depends on S*i = . . . (11
Ili -1
a) the practice-related objective of the calibration.
with the mean
The working range shall cover, as far as possible, the 10
application range for water, waste water, and sludge analy-
sis. The most frequently expected Sample concentration c Y-‘,J *
.- j=l
should lie in the centre of the working range. Yj = fori = 1 ori = 10 . . . (2)
yli
b) feasibilities of technical realizability.
The variances are tested (F-test) for significant differentes at
The measured values obtained must be linearly correlated to the limits of the working rangeL5t 6].
the concentrations. This requires that the measured values
obtained near the lower limit of the working range tan be The test value PG is determined for the F-test from
distinguished from the blanks of the method. The lower equation (3).
limit of the working range should therefore be equal to or
greater than the limit of detection of the method. Dilution
and concentrating Steps should be feasible without the risk PG = 2 forsyo > sf . . . (3)
of bias. 1

c) the variance of the information values must be indepen- S*

dent of the concentration. PG = --$ for s: > sf0
10

The independence is verified by a statistical test on the

PG is compared with the tabled values of the F-distribution[?
linearity16t *l.
Decision :
4.1.1 Reparation of the calibration
a) If PG < Ffl. I f2. I 099
I the differente between the
After establishing the preliminary working range, measured variances sf and SS is not significant.
values of at least five (recommended N = 10) Standard
samples are determined. The concentrations, xj, of these stan-
dard samples shall be distributed equidistantly over the working
b) If PG > q, ; j-2; 0,99 the differente between the

range. In Order to check for the homogeneity of the variances, variances sf and s$ is significant.
ten replicates of each of the lowest and the highest concentra-
tions (x, and x,~) of the working range are determined. Ten in- If the differente between the variances is significant, the
formation values, )pi jl result from these series of measurements preliminary working range should be made smaller until the dif-
(see table 1). ’ ference between the variances is found to be random only.

Table 1 - Data sheet for the calibration

Licensed to LAAM /MANUEL SOARES

ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
3
lSO,8466-1 : 199O’E)

4.13 Test for linearity[*J ‘# *] The measurement against a blank is not allowed, since thereby
valuable information on the magnitude of the blank will be lost.
The easiest test for the linearity is the graphical representation The comparison medium for zeroing the instrument is always, if
of the calibration data with the calculated regression line. Any possible, a pure solvent (e.g. pure water).
unlinearity is evident (see figure 2).

5 . Table 2 - Data sheet for simple linear regression

B
c
.-0
06I i xj
I
j
lI
x; -vj -V2i Xj’uj

051 I

z-
+E
8
04J
--y-----------l-----
1

g:ppIIj
3
03 J
-- ---- --L------

4 -t- , t ----t---l----l

5 .-----j----~----~--q-----j

\ , , , ., , ; . ,)c
~--
6
7
.--~~ --L-----
1 t------t-m
I
I

--_.- -----+------ ---

12345-6789
Concentration

Figure 2 - Graphical linearity check

In the statistical linearity test the calibration data are used to

calculate a linear calibration function as weil as a non-tin.ear
calibration function, both with the residual Standard deviation
Syl Or 52-
The ten data Sets, consisting of the va2ues of Xi and Yj, are sub-
The differente of the variances OS* is calculated from equation mitted to a linear regression analysis to obtain the coefficients a
(4) .. and b of the calibration function whsich describe the linear cor-
relation between the concentration x as an independent
OS’ = (N - 2) s;, - (IV - 3) sy, . . . (4) variable, and the measured value y as a dependent variable.

Degrees of freedom : f = 1. The calibration function as well as the characteristics of the

method should result from data obtained from a working range
OS* and the variance of the non-linear calibration function sv2 x1 to xIo, as received from the measurement and not corrected
for blanks. Generally, no blank value (concentration x = 0) is
are submitted to a F-test in Order to examine for significant dif-
to be included in the calibration experiment and, consequently,
ferences.
in the least-squares fit of the regression.
The test value PG required for the F-test is calculated from
equation (5) The linear calibration function is given by equation (6)

OS* Y =U+bX . . . (6)

PG =- . . . (5)
S*
Y*
The coefficients are obtained from equations (7) for sensitivity
Decision : (slope of the calibration function) and (8) for the Ordinate
intercept (calculated blank)
a) If PG 4 F: The non-linear calibration function does not
lead to a significantly better adjustment, e.g. the calibration N
function is linear.
(Xj - XI ' (Yj - Y)
c
b) If PG > F: The working range should be reduced as far i=l
b = p--_Ip--m_- . . . (7)
as possible to receive a linear calibration function; otherwise N
the information values of analyzed samples must be (x; - XI*
evaluated using the non-linear calibration function.
Y
M’
.
I= 1

4.2 Calibration and characteristics of the method

a = r--b% . . . (8)
After the final working range is established, ten Standard
samples are analyzed in accordance with all the Steps of the The coefficients provide an estimate of the true function, which
analytical method in Order to obtain ten (IV = 10) measured is limited by the unavoidable procedural scatter. The precision
values Yj (see table 2). of the estimate is quantified by the residual Standard deviation,
Licensed to LAAM /MANUEL SOARES
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
4
 ISO 8466-1 : 1990 (EI

sv, which is a measure of the scatter of the information values Fromsthe law of error propagation it follows that, for each value
about the calibration line and is given by equation (9). X, a confidence interval for the true value y exists whose
limiting Points are on two hyperbolic paths bracketing the
calibration line. Between these paths the true calibration func-
N
tion tan be expected with a significance level of a (fl : N - 2,
[Y j - (a + bx,)l*
c i=l
confidence level = 1 - a), determined by Student’s t-factor.
(9)
N-2 The confidence intervals for analytical results, calculated from
the calibration function, are given by the intersections with the
respective hyperbolic paths in figure 3. The estimation of the
4.3 Assessment
confidence intervals are given by equation (12) [71
The concentration of an analyzed Sample is obtained
7,,* = % + VB (.%,
a) from the measured valuey, to give ?
.a
Y -a Y -a
x =- . . . (IO) .2,,2 = b
b

or

b) from the mean of a series of replicates, 7, petformed on

(jf- Fl2 (12)
N
the same original Sample, to give ? b*x (Xi -
.
\ Z= 1 /
7-a
g=- . . . (11) A

b NOTE - If i; = 1, F,,* = X1,2-

As to the uncertainty of an analytical result, keep in mind that Equation (12) indicates that the confidence interval VB(?)
the analytical error is a combination of the uncertainty of the brackets the true analytical value with a range governed by the
determination of the measured value, and the uncertainty of statistical security of Student’s distribution. The magnitude of
the estimation of the regression coeff icients[*! VB(g) is mainly determined by the number of replicates n^ and

0
7
0 I
I
I
I
/ 1 I
a /’ I I
- vl3m
I I
I I
I
I
- 4-
/ 4 - X 10
Working range

Figure 3 - Working range x, to xlo, calibration line with confidence band and a Single analytical result with its
appertaining confidence interval
Licensed to LAAM /MANUEL SOARES
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
5
ISO 8466-1 : 1990 (E)

their results, the mean of the information values 7, as weil as 5.1 Choice of working range
the characteristics of the method, the residual Standard devi-
ation s,,, and the sensitivity b. For the analysis of drinking water and surface water, a working
range of 0,05 mg to 0,5 mg (NO;)/1 is appropriate.
The quality of the analytical procedure increases therefore with
increasing sensitivity and decreasing residual Standard deviation. 5.1.1 Testing the hornogeneity sf the variancesl)
The Standard deviation of the method sXO[see equation (1311 is
the characteristic which allows the analyst to check the quality According to the approach outlined in 4.1 .l, the variances ~2
of his own work. of the information values obtained from the Standard concen:
trations at the lower or upper limit of the working range respec-
sXO =2 . . . (13) tively, were determined (sec table 4). ’
b
The test value PG for the F-test is calculated from equation (3)
For the comparison of different standardized analytical
methods, the coefficient of Variation of the method, expressed S2 13,54 x IO-6
as a percentage, is given by equation (14) = 2,9
PG=s= 1 4,67 x 10-6

v xo = s~xloo . . . (14) Consulting the F-tablesL5] forfi = f2 = YE- 1 = 9 degrees of

freedom for the variances ST and S$ gives

5 Example F(9,9; 0,991 = 5,35

The photometric determination of nitrite is used to demonstrate The comparison of the calculated value PC with the tabled one
the calibration and the subsequent estimation of the statistical indicates a random differente between the variances under
characteristics of the method and their influence on the final examination. As the variances are homogeneous a simple
results of the evaluation. regression analysis may be petformed.

Table 3 - Data sheet for the calibration of NO;

i xi
mgll Y*l,l Yi,2 Yi,3 Y-1,4 Y-lt5 ’ Yfj
i, Y-1,7 i, R,a , &,!3 j Yi, 10

1 0,05 0,140 0,143 0,143 0,146 0,144 0,145 0,344 ' 0,146 0,145 0,148
2 0,lO 0,281
1
3 0,15 0,405 j
I
4 0,20 0,535 /
I
5 0,25 0,662 )
I
6 0,30 0,789 1
I
7 0,35 0,916 1
1
8 0,40 1,058
9 0,45 1,173
10 =N 0,50 1,303 1,302 1,300 1,304 1,300 1,296 1,295 1,301 1,296 1,306

Table 4 - Data sheet for the analysis sf variance. Object: nitrite

I

i xi Y. Yi,2 Yi,3 Y* Yi, 5 yi,6 Y* yi,8 Yi, 9 Yi, 10 s?

mgll Ex% Ext.*) Ext.*) Ex;.:) Ext.*) Ext. *) EX:,’ ) Ext.“, Ext. *) Ext. *) mg& 12
1 0,05 0,140 0,143 0,143 0,146 0,144 0,145 0,144 0,146 0,145 0,148 4,67. IO-6
I 10 1 0,50 1 1,303 / 1,302 1 1,300 1 1,304 1 1,300 / 1,296 j 1,295 1 1,301 1 1,296 1 1,306 11356 . lW6 1
I *c) Ext. : (extinction) I

1) Forthe sake oftransparency, alldimensions have intentionally been omitted in allequations, withoutambiguity. The dimensions are finallyadded
to the result.
Licensed to LAAM /MANUEL SOARES
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
6
5.1.2 Testing linearity The Standard deviation of the method is calculated from
equation ( 13)
A non-linear regression function may be derived [*l as

Y = 0,013 5 + 2,62 x - 0,081 8 ~2, giving a residual stan- s

xo
= sy
b-
- 0,005
-
2,575
2
2
= 0,002 0
dard deviation of sY2 = 0,005 2 mg/l.

The residual Standard deviations of the linear and the non-linear The coefficient of Variation of the method, expressed as a
calibration function, sY, and su2, are compared: percentage, is given by equation (14)

= 0,005 2 mg/1 (see 5.2 for calculation procedure) Sxo 0,002o x 100
sYl V xo = x x loo= = 0,73
0,275
sY*
= 0,005 2 mg/l.

As both residual Standard deviations are equal, the differente 5.3 Evaluation
of the variances DS* [see equation (411 does not need to be
5.3.1 Single determination
calculated. The non-linear calibration function does not lead to
a significantly better adjustment, e.g. the calibration function is The analysis of an unknown Sample, performed in the same
linear. way as the analysis of the Standards, gave an information value
jk 0,641 (extinction). The analytical result was obtained using
5.2 Calibration and characteristics of the method equation (12) with a confidence interva195 %, t(8; 0,951 = 2,31

Since the prerequisites for the Performance of a simple linear 0,641 - 0,018
regression are fulfilled, the calibration function and the %,2 = +
2,575
characteristics of the method tan be calculated using equations
(71, (8) and (13). The results are shown in table 5.
(0,641 - 0,726 2)*
+ 0,002 0 x 2,31 - + - +
The slope, b, as a measure for the sensitivity, is calculated from (2,575)* x 0,206 25
equation (7)
= (0,242 + 0,005) mg/1
N
Thus the true value of the concentration tan be expected
(Xi - 33 ’ (yi - jd
c within the range 0,237 < x < 0,247 mg/l, with a confidence
i=l
b = = 2,575 level of 0,95.
N

(X i- 32 5.3.2 Replicate analysis

c
i=l
For three replicate determinations, the analytical method gave
The Ordinate intercept, a, (calculated blank) is calculated from the information values 0,641; 0,631 and 0,633.
equation (8)
Calculation of the analytical result:
a = v-- b X = 0,726 2 - 2 1575 2 x 0 1275 = 0 I 018 Ext .
0,635 - 0,018
%,2 = +
The residual Standard deviation, sY, is calculated from 2,575
equation (9).

(Y i - yi)*
- +
(0,635 - 0,726 2)*
(2,575)* x 0,206 25
sy = = 0,005 2 [Ext.l
N-2 = (0,240 + 0,003) mg/l

The equation for the straight line is given by equation (4) Thus the true value of the concentration tan be expected
within the range 0,237 < x < 0,243 mg/I, with a confidence
Y = 0,018 + 2,575 2 x level of 0,95.

Table 5 - Data sheet for the regression

4

1 1 1 $i 1 ,66, i
mg0
xi
dV*1

4 I 0,20 1 0,535
I 6

8
0,30

0,35

0,40
0,789

0,916

1,058
-
T--t--
--

r / 10
9

c I 2f75l
i=l I
0,45

0,50

I
1,173

1,303

7,262
I
l

X = 0,275 mg /I 7 = 0,726 (Extinction)

Licensed to LAAM /MANUEL SOARES
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
7
rSO8466-1 : 1990 (El

Annex A
(informative)

Bibliography

111 VONDERHEID, C., DAMMAN, V. , DÜRR, w., FUNK, W. and KRUTZ, H., Statistical methods a nd Performance characteristics for the
assessment and comparison of analytical procedures. An approach to standardization, Vom Wasser 57, (1981), pp. 59-74.

[Zl MANDEL, J., The statistical analysis of experimental data, lnterscience PubL, J. Wiley & Sons, (1964), New York.

[31 GOTTSCHALK, G., Standardization of quantitative analytical procedures, Z Ana!. Chem. 275, (1975), pp. 1-10.

141 FRANKE, J.P., dE

tion spectrometry
ZEEW, R.A. and HAKKERT, R., Evaluation and optimization
and anodic stripping voltammetry,
of the Standard addition
AnaL Chem. 50, (1978), pp. 134-1380.
method for atomic absorp-

151 GRAF, U., HENNING, H.J. and STANGE, K., Formulae and tables of mathematical statistics, 2nd Edition, Springer Verlag, (1966),
Berlin, Heidelberg, New York.

BI SACHS, L., Methods for statistkaI evaluation, 3rd Edition, Springer Vdag, (1971), Berlin, Heidelberg, New York.

[71 BROWNLEE, K.A., Statistical theory and methodology in science and enghering, J. Wiley & Sens, (1965), New York.

BI DOERFFEL, K., Statistics in Chemical analysis, VEB- Verlag für die Grundstoffindustrie, (1966), Leipzig.

191 WAGNER, R., Evaluation of BOD-dilution series - Presentation of a Computer program, Vom Wasser 58, (1982), pp. 231-255.

no1 Commissariat a I’energie atomique, Statistique appliquee & l’exploitation des mesures, tome 2, Masson, (1978), pp. 345-379.

UDC 543.3 : 53.088

Descriptors: water, quality, tests, Chemical analysis, calibration, statistical analysis.

Price based on 8 pages

Licensed to LAAM /MANUEL SOARES
ISO Store order #:913756/Downloaded:2008-05-21
Single user licence only, copying and networking prohibited
8