If I follow, your dependent (or outcome) variable is "Sensory Processing Sensitivity". If so, how is it measured? Thanks for clarifying.

Hi @Bruce Weaver. Thanks for your reply. Yes Sensory Processing Sensitivity is the DV. It is measured continuously.

Hello Tadgh,

I'd suggest trying bootstrap/resampling solution; this approach frees you from having to worry about distributional assumptions associated with model errors (residuals).

As for the transformation approach, the way(s) in which a distribution departs from the target (for you, normal) distribution govern what then would be plausible choices for transformations. For example, a negative skew is exacerbated by applying log or root transforms. (Apologies if you already know this.) As well, the choice to jettison data points as outliers when the values are not demonstrably bogus is a tough one. One option is to run solutions with and without the "offending" cases and report both sets of results.

Good luck with your work.

Hi Tadgh Connery. I should have been clearer with my question. Is sensitivity a percentage, for example? Or is it on some other kind of scale? If it is a percentage (or proportion), then you might be better off using beta regression or a fractional regression model. What software do you have access to, by the way? HTH.

Hi Bruce Weaver .. Sorry I should have been clearer with my answer! It is measured on a scale. The software I am using is SPSS.

Hi David Morse

Thanks for your reply! I will try running it again with a bootstrap solution and see what happens. Thank you also for your advice regarding transformations - I was not aware of the direct implications that they have on distribution.

Thanks again!