8.6 Types of Measurement Errors
125
Random measurement error: a component of the measurement error that in
replicate measurements varies in an unpredictable manner.
Clause 2.19; ISO/IEC Guide 99
8.7 Systematic Errors
Systematic errors can stem from:
– The work of the analysts whose experience and practice result in a specific procedure;
– Incorrect performance or adjustment of the measuring device (e.g., weight, measuring glassware, spectrometer);
– The feature of the analytic method (e.g., effectiveness of extraction, an incomplete
run of the reaction).
Systematic errors cause a systematic bias of results and can cause their overestimation (positive errors) or underestimation (negative errors). Systematic errors, due
to their nature (are constant in specified conditions) should be eliminated, if possible,
from the measurement result. Of course, this is conditional upon the fact that their
value can be determined.
In cases when it is possible to determine the value of the systematic error, for
example, through comparison of the result obtained by a given method with the
result obtained through the reference method, or through the analysis of reference
material, it is possible to compensate for that value of the error in the measurement
result. Such a procedure is valid when it is assumed that the systematic error has been
determined correctly—that is that the assigned uncertainty is small, compared with
the bias. That assumption would be justified if the systematic error was determined by
using an infinite number of measuring results. Such a simplification can be applied in
cases when it can be demonstrated that the uncertainty of determining the correction
is very small in comparison to the observed spread of the results. In practice, the
component connected with the correction dominates and is an essential component
of the measurement uncertainty.
Systematic measurement error belongs to the category of influences that in metrological practice occur as correction or errors of the indications of the measuring
devices. They are characterized by a value accompanied with associated uncertainty.
In direct measurements, the measurement result is usually adjusted for the value
of those systematic influences, and the measurement uncertainty includes only the
random effects. Another procedure is also possible: namely, including the whole systematic effect into the confidence interval of the measurement result, and, therefore,
treating it as a component of uncertainty. This is especially beneficial in indirect measurements, in which the appropriate adjustment for the correction value can change
the definition of the measured quantity itself.
125
Random measurement error: a component of the measurement error that in
replicate measurements varies in an unpredictable manner.
Clause 2.19; ISO/IEC Guide 99
8.7 Systematic Errors
Systematic errors can stem from:
– The work of the analysts whose experience and practice result in a specific procedure;
– Incorrect performance or adjustment of the measuring device (e.g., weight, measuring glassware, spectrometer);
– The feature of the analytic method (e.g., effectiveness of extraction, an incomplete
run of the reaction).
Systematic errors cause a systematic bias of results and can cause their overestimation (positive errors) or underestimation (negative errors). Systematic errors, due
to their nature (are constant in specified conditions) should be eliminated, if possible,
from the measurement result. Of course, this is conditional upon the fact that their
value can be determined.
In cases when it is possible to determine the value of the systematic error, for
example, through comparison of the result obtained by a given method with the
result obtained through the reference method, or through the analysis of reference
material, it is possible to compensate for that value of the error in the measurement
result. Such a procedure is valid when it is assumed that the systematic error has been
determined correctly—that is that the assigned uncertainty is small, compared with
the bias. That assumption would be justified if the systematic error was determined by
using an infinite number of measuring results. Such a simplification can be applied in
cases when it can be demonstrated that the uncertainty of determining the correction
is very small in comparison to the observed spread of the results. In practice, the
component connected with the correction dominates and is an essential component
of the measurement uncertainty.
Systematic measurement error belongs to the category of influences that in metrological practice occur as correction or errors of the indications of the measuring
devices. They are characterized by a value accompanied with associated uncertainty.
In direct measurements, the measurement result is usually adjusted for the value
of those systematic influences, and the measurement uncertainty includes only the
random effects. Another procedure is also possible: namely, including the whole systematic effect into the confidence interval of the measurement result, and, therefore,
treating it as a component of uncertainty. This is especially beneficial in indirect measurements, in which the appropriate adjustment for the correction value can change
the definition of the measured quantity itself.