4.3 Preliminary Operations
157
compromise between the number of samples (size) to be processed and
the costs/efforts incurred. Ideally, one should use infinite portions of the
sample, which, however, is an unrealistic approach. In any case, the sampling
plan should use the minimum number of samples that will ensure representativeness at the intended level. Statistics (chemometrics) provides invaluable
technical support for this purpose. The ultimate goal is to maximize representativeness and precision with the smallest possible number of samples and
sample size.
Depending on the global approach adopted in the plan, sampling can be of
one of the following types:
(a) Intuitive. The analyst, by personal choice, selects the object portion to
be processed (e. g. when a texture or colour change is observed in the
studied material or some isolated alteration in a production process is
suspected).
(b) Statistical. The selection is based on statistical rules that are applied after the
analytical problem has been defined (e.g. when the minimum number of
samples to be used is determined on the assumption of a Gaussian distribution in the object composition, characterized by a standard deviation, and in
compliance with the accepted levels of error in the results and standard
deviation in the measurements).
(c) Directed. This is the preferred choice when the analytical problem demands
special information (e. g. when, in analysing natural water, the concentration
of metal species in suspended particles of a specific size is the only parameter needed).
(d) Protocol-based. By legal mandate or by the client's demand, the sampling
procedure to be used in some instances is one regulated by a standard or
specified in an official publication.
Notwithstanding this classification, two or more of these sampling types are
virtually indistinguishable in practice.
Broadly speaking, the sample is a portion of the object that contains the
measurand or analyte; all are parts of the ranking discussed in Sect. 1.6. This
definition of sample is impractical and, especially, oversimplified. One way of
making it more explicit is by using different qualifiers to avoid confusion. The
two approaches summarized in Fig. 4.11 illustrate the usage of the word
"sample" in Analytical Chemistry.
Depending on the way it is collected, a sample can be qualified with different
attributes. Thus, a random sample is one selected in such a way that any portion
of the object will have had a given likelihood (e. g. 95 %) of being withdrawn. A
representative sample is one selected by applying a sampling plan consistent
with the definition of the analytical problem addressed. A selective sample is a
sample collected by following a selective sampling procedure. A stratified
sample is one randomly chosen from a previously collected portion, stratum or
well-defined zone of the object. Finally, a convenience sample is one selected in
terms of accessibility, cost-effectiveness, efficiency, etc. (i. e. because it possesses
favourable sampling characteristics).
157
compromise between the number of samples (size) to be processed and
the costs/efforts incurred. Ideally, one should use infinite portions of the
sample, which, however, is an unrealistic approach. In any case, the sampling
plan should use the minimum number of samples that will ensure representativeness at the intended level. Statistics (chemometrics) provides invaluable
technical support for this purpose. The ultimate goal is to maximize representativeness and precision with the smallest possible number of samples and
sample size.
Depending on the global approach adopted in the plan, sampling can be of
one of the following types:
(a) Intuitive. The analyst, by personal choice, selects the object portion to
be processed (e. g. when a texture or colour change is observed in the
studied material or some isolated alteration in a production process is
suspected).
(b) Statistical. The selection is based on statistical rules that are applied after the
analytical problem has been defined (e.g. when the minimum number of
samples to be used is determined on the assumption of a Gaussian distribution in the object composition, characterized by a standard deviation, and in
compliance with the accepted levels of error in the results and standard
deviation in the measurements).
(c) Directed. This is the preferred choice when the analytical problem demands
special information (e. g. when, in analysing natural water, the concentration
of metal species in suspended particles of a specific size is the only parameter needed).
(d) Protocol-based. By legal mandate or by the client's demand, the sampling
procedure to be used in some instances is one regulated by a standard or
specified in an official publication.
Notwithstanding this classification, two or more of these sampling types are
virtually indistinguishable in practice.
Broadly speaking, the sample is a portion of the object that contains the
measurand or analyte; all are parts of the ranking discussed in Sect. 1.6. This
definition of sample is impractical and, especially, oversimplified. One way of
making it more explicit is by using different qualifiers to avoid confusion. The
two approaches summarized in Fig. 4.11 illustrate the usage of the word
"sample" in Analytical Chemistry.
Depending on the way it is collected, a sample can be qualified with different
attributes. Thus, a random sample is one selected in such a way that any portion
of the object will have had a given likelihood (e. g. 95 %) of being withdrawn. A
representative sample is one selected by applying a sampling plan consistent
with the definition of the analytical problem addressed. A selective sample is a
sample collected by following a selective sampling procedure. A stratified
sample is one randomly chosen from a previously collected portion, stratum or
well-defined zone of the object. Finally, a convenience sample is one selected in
terms of accessibility, cost-effectiveness, efficiency, etc. (i. e. because it possesses
favourable sampling characteristics).