Time Dependence of Data

When measuring the variability in the data you should also consider that a part of the variation is caused by time (due to ageing, or periodic demands, etc). Thus looking at the distribution alone may give the wrong impression. The following  interactive example  shows how the distribution depends on the time slot which is used to analyse a signal. You may also go to the  DataLab  to experiment with some time series yourself.

Time series have the advantage that we normally know from the experimental conditions that the data is a time series. A similar (or even worse) situation may occur if some measured data depends on another variable. Here the distribution of the data will not reveal anything. Even plotting the data against the index will not reveal any dependency. The variable(s) causing the dependence may sometimes be found by trial and error guided by speculation and experience. Some help can come from scatter plots.