Hello Johann,

I've noticed that some graph theory studies use parametric statistics and others non-parametric. I am not sure how they checked for normality of graph metrics. Is there a way in GraphVar to check the distribution of the graph metrics in order to determine whether to run parametric or non-parametric statistics?

Thanks in advance!

Johann Kruschwitz

Mon, 02/24/2020 - 13:06

Permalink

## distribution of graph-metrics

Hi,

there is currently no option to check the distribution of graph metrics in GraphVar itself. However, you could export them and examine these e.g. in SPSS. In my opinion, if the number of repetitions is high enough, the usage of non-parametric tests (e.g. permutation testing) is a better option then simply using pre-defined assumptions about the distribution of thedata (as in parametric tests). This is because the non-parametric test-statistical distribution will be directly tailored to your data and therefore in most cases also be more accurate. As GraphVar runs also the parametirc test in parallel when selecting to calculate non-parametric tests, you can check in the results viewer if effects are different between the two approaches (drop down window: parametic vs. permuation).

Best,

Johann