The basis of all statistical analyses is the data. The data set to be
analyzed should describe one or more than one characteristic of the observed
object. When analyzing data, we should always be aware of the data source,
keeping the following points in mind:
Details of the data acquisition: the type of data acquisition (manual,
automatic, frequency and intensity resolution, sampling plan, etc.) may
influence the results considerably, and may require different methods of
Is the measured data a population or a sample? Different
formulas and the precision of results follow from this.
What is the precision of the measurements. This knowledge is necessary
to avoid an excessive number of decimal places. It may be misleading, for
example, if the results of an exit poll during elections are presented
as "23.71% for party xy" if the uncertainty is approximately 2 %. A correct
specification would be 23 +/- 2 %.
Is the observed variable really significant for the problem to be
In some cases, the selection of the significant variables
is not obvious
What exactly is the problem to be solved? It can
be shown, that using a set of statistical methods by trial and error (especially
test statistics) increases the probability of drawing certain conclusions.