28
methods have been developed without altering the underlying signal or signal characteristics. Most often, these data filtering methods are based on a statistical model
to determine the so-called dynamic mean. Another useful model is the SavitzkyGolay method. This model is based on the principle of inserting a dynamic mean
into a polynomial arrangement. Basically, such models represent the normalization
of the share of measuring points (carrier variables), where more weight is given to
the central data and less to the peripheral ones.
Environmental investigations deal with numerous data. For a large number of
variables with high variance in distribution and high standard deviation (which is
often the case for the distribution of environmental variables), there is a problem
treating all results simultaneously. In such cases, the variables with the largest variance will have the greatest impact on the outcome of the procedure applied (the
appropriate statistical method).
Data transformation and/or standardization is used as a possible solution to the
problem. Usually, in asymmetric distributions, the results are first transformed using
a logarithmic transformation that helps to facilitate homogeneity in the variance.
This procedure will emphasize the influence of the variables with high variance. If
this is not the desired effect of the transformed data, then standardization is
performed.
Transforming data means performing the same mathematical operations on each
component of the original data. If the original data is multiplied or divided by a
specific coefficient or is repeatedly subtracted or added, then we are talking about
linear transformations. But these linear transformations do not change the shape of
the data (i.e., their distribution) and therefore do not help normalize the distribution
of the data.
It is usually characteristic of data on certain environmental variables whose distributions are log-normal or positively curved. Therefore, it is necessary, before
proceeding with their processing, to undergo a transformation first, that is, to
normalize their distributions. Logarithmic transformation is widely applied in order
to normalize positively curved data distributions. The purpose of applying a particular data transformation is to actually reduce the difference between extreme values.
The Box-Cox transformation method was also used to normalize the data (Box and
Cox 1964).
Regression analysis is a typical example of a statistical tool that can be widely
applied in solving heterogeneous and multiparametric distributions. In the determination of the correlation parameters, for defining the appropriate model, the regression analysis method can be successfully used. It should be borne in mind that
regression analysis can be expressed with other dependencies, not just linear dependencies, as well as multiple parametric dependencies.
Factor analysis is an interdependence technique because it looks for a group of
variables that are similar in that they “move together” and therefore have great interdependence. When one variable has a large value, then the other variables in the
group have a large value. For the effective application of factor analysis, as well as
other multivariate interdependence techniques, it is necessary to have a minimal
B. Balabanova
methods have been developed without altering the underlying signal or signal characteristics. Most often, these data filtering methods are based on a statistical model
to determine the so-called dynamic mean. Another useful model is the SavitzkyGolay method. This model is based on the principle of inserting a dynamic mean
into a polynomial arrangement. Basically, such models represent the normalization
of the share of measuring points (carrier variables), where more weight is given to
the central data and less to the peripheral ones.
Environmental investigations deal with numerous data. For a large number of
variables with high variance in distribution and high standard deviation (which is
often the case for the distribution of environmental variables), there is a problem
treating all results simultaneously. In such cases, the variables with the largest variance will have the greatest impact on the outcome of the procedure applied (the
appropriate statistical method).
Data transformation and/or standardization is used as a possible solution to the
problem. Usually, in asymmetric distributions, the results are first transformed using
a logarithmic transformation that helps to facilitate homogeneity in the variance.
This procedure will emphasize the influence of the variables with high variance. If
this is not the desired effect of the transformed data, then standardization is
performed.
Transforming data means performing the same mathematical operations on each
component of the original data. If the original data is multiplied or divided by a
specific coefficient or is repeatedly subtracted or added, then we are talking about
linear transformations. But these linear transformations do not change the shape of
the data (i.e., their distribution) and therefore do not help normalize the distribution
of the data.
It is usually characteristic of data on certain environmental variables whose distributions are log-normal or positively curved. Therefore, it is necessary, before
proceeding with their processing, to undergo a transformation first, that is, to
normalize their distributions. Logarithmic transformation is widely applied in order
to normalize positively curved data distributions. The purpose of applying a particular data transformation is to actually reduce the difference between extreme values.
The Box-Cox transformation method was also used to normalize the data (Box and
Cox 1964).
Regression analysis is a typical example of a statistical tool that can be widely
applied in solving heterogeneous and multiparametric distributions. In the determination of the correlation parameters, for defining the appropriate model, the regression analysis method can be successfully used. It should be borne in mind that
regression analysis can be expressed with other dependencies, not just linear dependencies, as well as multiple parametric dependencies.
Factor analysis is an interdependence technique because it looks for a group of
variables that are similar in that they “move together” and therefore have great interdependence. When one variable has a large value, then the other variables in the
group have a large value. For the effective application of factor analysis, as well as
other multivariate interdependence techniques, it is necessary to have a minimal
B. Balabanova