I have two non-normal variables (one DV, one IV) and a few 7-point Likert scale IVs (normally distributed). The non-normal variables are centrality scores from network analysis - DV is from the collaboration matrix; IV is from the friendship matrix. Transforming the data did not work out - unfortunately.

Here is what I would like to test: DV: centrality score - collaboration IV: centrality score - friendship annual budget number of employees years of operation

I think I have to go with a non-parametric test. The problem - at least for me - is that the 2 non-normal variables are continuous and have different ranges.

I thought about assigning a rank instead of the centrality scores, but I am not sure which test to use.

I have the feeling I am overlooking something quite simple.

  • 2
    $\begingroup$ In spite of your assertion of the contrary, Likert scales are clearly not normal. Your 'here is what I would like to test' is unclear. When you put a '-' between two variable names are you trying to test some form of equality of location or something else? Could you clarify what you're trying to achieve in ordinary English? $\endgroup$
    – Glen_b
    Mar 10, 2013 at 2:22
  • 2
    $\begingroup$ The simple thing you're overlooking is that none of these variables is required to be normal. Only the residuals need to be normal (see: what-if-residuals-are-normally-distributed-but-y-is-not), & TBH, that assumption is the least important assumption. $\endgroup$ Mar 16, 2013 at 4:50