46
2 Analytical Properties
Table 2.1. Types of errors in Analytical Chemistry, characterized according to various criteria
DIFfERENTIAL
TYPES OF
Relative
Analytical
ERROS
Reference used
SOURCE
for definition
SIGN
magnitude
property
(as a rule)
affected
variable
PRECISION
RANDOM
f-+
Random
Mean of a set
@
small
(specific
(indeterminate) •
uncertainty)
SYSTEMATIC
unique
small
(determinate)
......
Well-defined
Xl 6:)
8 jffltive
ACCURACY
alterations
(traceability)
or
GROSS
......
8 1l E9!1M'
large
the same sample or, simply, when the same measurement is made several times.
The reference for calculation of this type of error is the arithmetic mean ()(, p' }
for a series of determinations. Random errors vary in magnitude, which is rarely
very large. They result in overestimation or underestimation, depending on
whether they lead to data above or below the mean. Consequently, this type of
error is meaningless for an individual datum (result) unless a data set is available for comparison. It can be ascribed to individual results or to a method
(means) and is characterized by using statistical principles based on the normal
or Gaussian distribution (see Box 2.2). Random errors are the basis for the
analytical property precision and hence for specific uncertainty.
Box 2.2
Principal statistical landmarks for handling analytical chemical data
• The statistics applied to analytical chemical data is a pa rt of the multidisciplinary entity
called CHEMOMETRICS, which aims to expand and enhance analytical chem ical information by using less material, time and human resources (with the aid of computers) .
• Statistics is a key support for analytical properties in general and the following in
particular:
- Representativeness (sample collection rules)
- Accuracy (traceability)
- Precision (uncertainty)
- Sensitivity (definition of "limit of detection" and "limit of quantitation" via statistical
parameters).
2 Analytical Properties
Table 2.1. Types of errors in Analytical Chemistry, characterized according to various criteria
DIFfERENTIAL
TYPES OF
Relative
Analytical
ERROS
Reference used
SOURCE
for definition
SIGN
magnitude
property
(as a rule)
affected
variable
PRECISION
RANDOM
f-+
Random
Mean of a set
@
small
(specific
(indeterminate) •
uncertainty)
SYSTEMATIC
unique
small
(determinate)
......
Well-defined
Xl 6:)
8 jffltive
ACCURACY
alterations
(traceability)
or
GROSS
......
8 1l E9!1M'
large
the same sample or, simply, when the same measurement is made several times.
The reference for calculation of this type of error is the arithmetic mean ()(, p' }
for a series of determinations. Random errors vary in magnitude, which is rarely
very large. They result in overestimation or underestimation, depending on
whether they lead to data above or below the mean. Consequently, this type of
error is meaningless for an individual datum (result) unless a data set is available for comparison. It can be ascribed to individual results or to a method
(means) and is characterized by using statistical principles based on the normal
or Gaussian distribution (see Box 2.2). Random errors are the basis for the
analytical property precision and hence for specific uncertainty.
Box 2.2
Principal statistical landmarks for handling analytical chemical data
• The statistics applied to analytical chemical data is a pa rt of the multidisciplinary entity
called CHEMOMETRICS, which aims to expand and enhance analytical chem ical information by using less material, time and human resources (with the aid of computers) .
• Statistics is a key support for analytical properties in general and the following in
particular:
- Representativeness (sample collection rules)
- Accuracy (traceability)
- Precision (uncertainty)
- Sensitivity (definition of "limit of detection" and "limit of quantitation" via statistical
parameters).