Sampling,

Standardization,
and Calibration Chapter 8

Because a chemical analysis uses only a small fraction of the available sample, the process
of sampling is a very important operation. The fractions of the sandy and loam soil samples shown
in the photo that are collected for analyses must be representative of the bulk materials. Knowing
how much sample to collect and how to further subdivide the collected sample to obtain a laboratory
sample is vital in the analytical process. Sampling, standardization, and calibration are the focal
points of this chapter. Statistical methods are an integral part of all three of these operations.

I
© Bob Rowan; Progressive Image/CORBIS
n Chapter 1, we described a typical real-world analytical procedure consisting of several
­important steps. In any such procedure, the specific analytical method selected depends on
how much sample is available and how much analyte is present. We discuss here a general classi-
fication of the types of determinations based on these factors. After selecting the particular method
to be used, a representative sample must be acquired. In the sampling process, we make every
effort to select a small amount of material that accurately represents the bulk of the material being
analyzed. We use statistical methods to aid in the selection of a representative sample. Once the
analytical sample has been acquired, it must be processed in a dependable manner that main-
tains sample integrity without losing sample or introducing contaminants. Many laboratories use
the automated sample handling methods discussed here because they are reliable and cost effec-
tive. Because analytical methods are not absolute, results must be compared with those obtained
on standard materials of accurately known composition. Some methods require direct compari-
son with standards while others involve an indirect calibration procedure. Much of our discussion
­focuses on the details of standardization and calibration including the use of statistical procedures
to construct calibration models. We conclude this chapter with a discussion of the methods used to
compare analytical methods by using various performance criteria, called figures of merit.

8A Analytical Samples and Methods

Many factors are involved in the choice of a specific analytical method as discussed in
Section 1C-1. Among the most important factors are the amount of sample and the
concentration of the analyte.

8A-1 Types of Samples and Methods

Often, we distinguish a method of identifying chemical species, a qualitative analysis,
from one to determine the amount of a constituent, a quantitative analysis. Quan-
titative methods, as discussed in Section 1B are traditionally classified as gravimetric

Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
154  Chapter 8 Sampling, Standardization, and Calibration

methods, volumetric methods, or instrumental methods. Another way to distinguish

methods is based on the size of the sample and the level of the constituents.

Sample Size
As shown in Figure 8-1, the term macro analysis is used for samples whose
masses are greater than 0.1 g. A semimicro analysis is performed on samples
in the range of 0.01 to 0.1 g, and samples for a micro analysis are in the range
1024 to 1022 g. For samples whose mass is less than 1024 g, the term ultramicro
­analysis is sometimes used.
From the classification in Figure 8-1, we can see that the analysis of a 1-g sample
Sample Size Type of Analysis of soil for a suspected pollutant would be called a macro analysis whereas that of a
. 0.1 g Macro 5-mg sample of a powder suspected to be an illicit drug would be a micro analysis.
0.01 to 0.1 g Semimicro A typical analytical laboratory processes samples ranging from the macro to the
0.0001 to 0.01 g Micro ­micro and even to the ultramicro range. Techniques for handling very small samples
, 1024 g Ultramicro
are quite different from those for treating macro samples.

Constituent Types
The constituents determined in an analytical procedure can cover a huge range in
concentration. In some cases, analytical methods are used to determine major con-
stituents, which are those present in the range of 1 to 100% by mass. Many of the
gravimetric and some of the volumetric procedures discussed in Part III are exam-
ples of major constituent determinations. As shown in Figure 8-2, species present in
the range of 0.01 to 1% are usually termed minor constituents, while those present

Macro
Type of analysis

Semimicro

Micro

Ultramicro

Figure 8-1 Classification of analyses 0.0001 0.001 0.01 0.1

by sample size. Sample size, g

Major
Type of constituent

Minor

Trace

Ultratrace

Figure 8-2 Classification of 1 ppb 1 ppm 0.1 % 100 %

constituent types by analyte level. Analyte level

Unless otherwise noted, all content on this page is © Cengage Learning.

Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 8A Analytical Samples and Methods  155

in amounts between 100 ppm (0.01%) and 1 ppb are called trace constituents.
Components present in amounts lower than 1 ppb are usually considered to be ul-
tratrace constituents.
Determining Hg in the ppb to ppm range in a 1-mL (<1 mg) sample of river Analyte Level Type of Constituent
water would be a micro analysis of a trace constituent. Determinations of trace and
1 to 100% Major
ultratrace constituents are particularly demanding because of potential interferences 0.01 (100 ppm) to 1% Minor
and contaminations. In extreme cases, determinations must be performed in spe- 1 ppb to 100 ppm Trace
cial rooms that are kept meticulously clean, free from dust and other contaminants. , 1 ppb Ultratrace
A general problem in trace procedures is that the reliability of results usually decreases
dramatically with a decrease in analyte level. Figure 8-3 shows how the relative stan-
dard deviation between laboratories increases as the level of analyte decreases. At the
ultratrace level of 1 ppb, interlaboratory error (%RSD) is nearly 50%. At lower levels,
the error approaches 100%.

8A-2 Real Samples
The analysis of real samples is complicated by the presence of the sample matrix. The
matrix can contain species with chemical properties similar to the analyte. Matrix
components can react with the same reagents as the analyte, or they can cause an
instrument response that is not easily distinguished from the analyte. These effects
interfere with the determination of the analyte. If the interferences are caused by
extraneous species in the matrix, they are often called matrix effects. Such effects
can be induced not only by the sample itself but also by the reagents and solvents
used to prepare the samples for the determination. The composition of the matrix
containing the analyte may vary with time as is the case when materials lose water
by dehydration or undergo photochemical reactions during storage. We discuss ma-
trix effects and other interferences in the context of standardization and calibration
methods in Section 8D-3.
As discussed in Section 1C, samples are analyzed, but species or concentrations
are determined. Hence, we can correctly discuss the determination of glucose in blood
❮ Samples are analyzed, but
constituents or concentrations
are determined.
serum or the analysis of blood serum for glucose.

50
1 ppb

40
% Relative standard deviation

30
Figure 8-3 Interlaboratory error
as a function of analyte concentration.
20 Note that the relative standard
deviation ­dramatically increases as
1 ppm the analyte concentration decreases.
In the ultratrace range, the relative
10
standard deviation approaches 100%.
(Reprinted (adapted) with permission
0.01%
from W. Horowitz, Anal. Chem.,
1% 1982, 54, 67A–76A., DOI: 10.1021/
0
0 –2 –4 –6 –8 –10 ac00238a002. Copyright 1982
log concentration American Chemical Society.)
Unless otherwise noted, all content on this page is © Cengage Learning.

Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
156  Chapter 8 Sampling, Standardization, and Calibration

8B Sampling
A chemical analysis is most often performed on only a small fraction of the material
of interest, for example a few milliliters of water from a polluted lake. The composi-
tion of this fraction must reflect as closely as possible the average composition of
the bulk of the material if the results are to be meaningful. The process by which a
Sampling is often the most difficult representative fraction is acquired is termed sampling. Often, sampling is the most
aspect of an analysis. difficult step in the entire analytical process and the step that limits the accuracy of
the procedure. This statement is especially true when the material to be analyzed is a
large and inhomogeneous liquid, such as a lake, or an inhomogeneous solid, such as
an ore, a soil, or a piece of animal tissue.
Sampling for a chemical analysis necessarily requires the use of statistics be-
cause conclusions will be drawn about a much larger amount of material from
the analysis of a small laboratory sample. This is the same process that we dis-
cussed in Chapters 6 and 7 for examining a finite number of items drawn from
a population. From the observation of the sample, we use statistics, such as
the mean and standard deviation, to draw conclusions about the population.
The literature on sampling is extensive1; we provide only a brief introduction in
this section.

8B-1 Obtaining a Representative Sample

The sampling process must ensure that the items chosen are representative of
The composition of the gross
sample and the laboratory ❯ the bulk of material or population. The items chosen for analysis are often
called sampling units or sampling increments. For example, our population
sample must closely resemble the
average composition of the total might be 100 coins, and we might wish to know the average concentration of
mass of material to be analyzed. lead in the collection of coins. Our sample is to be composed of 5 coins. Each
coin is a sampling unit or increment. In the statistical sense, the sample corre-
In sampling, a sample population is sponds to several small parts taken from different parts of the bulk material. To
reduced in size to an amount of homo- avoid confusion, chemists usually call the collection of sampling units or incre-
geneous material that can be conve-
niently handled in the laboratory and ments, the gross sample.
whose composition is representative of For analysis in the laboratory, the gross sample is usually reduced in size and ho-
the population. mogenized to create the laboratory sample. In some cases, such as sampling pow-
ders, liquids, and gases, we do not have obvious discrete items. Such materials may
Identify the not be homogeneous because they may consist of microscopic particles of different
population
compositions or, in the case of fluids, zones where concentrations of the analyte
differ. With these materials, we can prepare a representative sample by taking our
­sample increments from different regions of the bulk material. Figure 8-4 illustrates
Collect a the three steps that are usually involved in obtaining the laboratory sample. Step 1 is
gross sample
often straightforward with the population being as diverse as a carton of bottles con-
taining vitamin tablets, a field of wheat, the brain of a rat, or the mud from a stretch
of river bottom. Steps 2 and 3 are seldom simple and may require tremendous effort
Reduce the gross and ingenuity.
sample to a
laboratory sample

Figure 8-4 Steps in obtaining a

laboratory sample. The laboratory 1
See, for example, J. L. Devore and N. R. Farnum, Applied Statistics for Engineers and Scientists,
sample consists of a few grams to at 2nd ed. Pacific Grove, CA: Duxbury Press, 2005, Ch. 4; J. C. Miller and J. N. Miller, Statistics
most a few hundred grams. It may and Chemometrics for Analytical Chemistry, 4th ed., Upper Saddle River, NJ: Prentice Hall, 2000;
constitute as little as 1 part in 107 B. W. Woodget and D. Cooper, Samples and Standards, London: Wiley, 1987; F. F. Pitard, Pierre
or 108 of the bulk material. Gy’s Sampling Theory and Sampling Practice, Boca Raton, Fl: CRC Press, 1989.

Unless otherwise noted, all content on this page is © Cengage Learning.

Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 8B Sampling  157

Statistically, the goals of the sampling process are:

1. To obtain a mean analyte concentration that is an unbiased estimate of the popu-

lation mean. This goal can be realized only if all members of the population have
an equal probability of being included in the sample.
2. To obtain a variance in the measured analyte concentration that is an unbiased
estimate of the population variance so that valid confidence limits can be found
for the mean, and various hypothesis tests can be applied. This goal can be
reached only if every possible sample is equally likely to be drawn.

Both goals require obtaining a random sample. Here the term random sample
does not imply that the samples are chosen in a haphazard manner. Instead a ran-
domization procedure is applied to obtain such a sample. For example, suppose our
sample is to consist of 10 pharmaceutical tablets to be drawn from 1000 tablets off a
production line. One way to ensure the sample is random is to choose the tablets to
be tested from a table of random numbers. These can be conveniently generated from
a random number table or from a spreadsheet as is shown in Figure 8-5. Here, we
would assign each of the tablets a number from 1 to 1000 and use the sorted random
numbers in column C of the spreadsheet to pick tablet 16, 33, 97, etc. for analysis.

8B-2 Sampling Uncertainties
In Chapter 5, we concluded that both systematic and random errors in analytical
data can be traced to instrument, method, and personal causes. Most systematic
errors can be eliminated by exercising care, by calibration, and by the proper use

Figure 8-5 Generating 10 random numbers from 1 to 1000 by a spreadsheet. The random

number function in Excel [=RAND()] generates random numbers between 0 and 1. The mul-
tiplier shown in the documentation ensures that the numbers generated in column B will be
between 1 and 1000. In order to obtain integer numbers, we right click on the selected cells
and choose Format Cells… from the drop down menu. We then choose Number and then
0 decimal places. So that the numbers do not change with every recalculation, the random
numbers in column B were copied and then pasted as values into column C using the Paste
Special… command on the Home ribbon. In column C the numbers were sorted in ascending
order using Excel’s Data Sort… command on the Data ribbon.

Unless otherwise noted, all content on this page is © Cengage Learning.

Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
158  Chapter 8 Sampling, Standardization, and Calibration
When sm # ss /3, there is no
point in trying to improve
the measurement precision.
Equation 8-1 shows that so will
be determined predominately by
the sampling uncertainty under
these conditions.
The gross sample is the
collection of individual sampling
units. It must be representative
of the whole in composition and
in particle-size distribution.
Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
of standards, blanks, and reference materials. Random errors, which are reflected in
the precision of data, can generally be kept at an acceptable level by close control
of the variables that influence the measurements. Errors due to invalid sampling are
unique in the sense that they are not controllable by the use of blanks and standards
or by closer control of experimental variables. For this reason, sampling errors are
ordinarily treated separately from the other uncertainties associated with an analysis.
For random and independent uncertainties, the overall standard deviation so for an
analytical measurement is related to the standard deviation of the sampling process ss
and to the standard deviation of the method sm by the relationship
s2o 5 s2s 1 s2m (8-1)
In many cases, the method variance will be known from replicate measurements of
a single laboratory sample. Under this circumstance, ss can be computed from mea-
surements of s0 for a series of laboratory samples, each of which is obtained from
several gross samples. An analysis of variance (see Section 7C) can reveal whether
the between samples variation (sampling plus measurement variance) is significantly
greater than the within samples variation (measurement variance).
❯ Youden has shown that, once the measurement uncertainty has been reduced to
one third or less of the sampling uncertainty (that is, sm # ss/3), further improvement
in the measurement uncertainty is fruitless.2 This result suggests that, if the sampling
uncertainty is large and cannot be improved, it is often a good idea to switch to a less
precise but faster method of analysis so that more samples can be analyzed in a given
length of time. Since the standard deviation of the mean is lower by a factor of !N ,
taking more samples can improve precision.
8B-3 The Gross Sample
❯ Ideally, the gross sample is a miniature replica of the entire mass of material to be
analyzed. It should correspond to the bulk material in chemical composition and in
particle-size distribution if the sample is composed of particles.
Size of the Gross Sample
For convenience and economy, the gross sample should be no larger than absolutely
necessary. Basically, gross sample size is determined by (1) the uncertainty that can
be tolerated between the composition of the gross sample and that of the whole,
(2) the degree of heterogeneity of the whole, and (3) the level of particle size at which
heterogeneity begins.3
The last point warrants amplification. A well-mixed, homogeneous solution of a
gas or liquid is heterogeneous only on the molecular scale, and the mass of the mol-
ecules themselves governs the minimum mass of the gross sample. A particulate solid,
such as an ore or a soil, represents the opposite situation. In such materials, the indi-
vidual pieces of solid differ from each other in composition. Heterogeneity develops
in particles that may have dimensions on the order of a centimeter or more and may
be several grams in mass. Intermediate between these extremes are colloidal materials
and solidified metals. With colloidal materials, heterogeneity is first encountered in
the range of 1025 cm or less. In an alloy, heterogeneity first occurs in the crystal grains.
2
W. J. Youden, J. Assoc. Off. Anal. Chem., 1981, 50, 1007.
3
For a paper on sample mass as a function of particle size, see G. H. Fricke, P. G. Mischler,
F. P. Staffieri, and C. L. Housmyer, Anal. Chem., 1987, 59, 1213, DOI: 10.1021/ac00135a030.