Learning by Simulations has been developed by Hans Lohninger to support both teachers and students in the process of knowledge transfer and acquisition . Click here for more information.

Share this Page: del.icio.us   Reddit   Facebook   StumbleUpon   MySpace   Digg   Twitter  

The correlation coefficient is one of the commonly used statistical measures. Despite its frequent use, the fact that the correlation coefficient shows a broad distribution if too few observations are used is not well-known.

This problem becomes worst when using only three or four observations for calculating the correlation. For example, if the actual correlation is zero, the correlation coefficient is distributed uniformly if only four observations are used (which means that the calculated correlation coefficient may in fact be any value between -1 and +1 with equal chances of occurrence).

English version [318 kB]
German version [318 kB]
After downloading please unpack all files of
the zipped packages and start the executable.
This program has been developed to show the broad distribution of the correlation coefficient when using only a few measurements. You can experiment with the distribution by changing the number of observations and the expected correlation coefficient.

Last Update: 2006-Jun-28