Hi, I'm trying to do my thesis and well, statistics is certainly not my thing but I'm trying so I hope I can finally finish my thesis after postponing it for like 2 years now (Please don't judge me). Anyway, I need help on some things. Firstly I'll explain what I have and what I have done and if anyone thinks that I have been thinking the wrong way, please help me out so that I don't do things the wrong way.
I have a dataset of river dolphin counts which I structured so that I have dolphins/km (so no longer count data). i did a normality test and found out that it is certainly not normal distribution, even after log-transforming the data. Thus, non-parametric tests were the option to go. I have two factors that I'm using for this, habitat (5 habitats) and locals (around 24), i.e. I have different aquatic habitats I surveyed to see if the dolphins are found more on some than others. I have also different "locals"/sites (bigger areas) that together comprise the whole study area. So in general I want to check how habitats and sites differ in terms of dolphins. I was a bit unsure about the test to use and thought about Kruskal-Wallis, however since I also wanted to see if the interaction of the factors habitat and locals was significant, I got to understand this was not possible with the Kruskal-Wallis. Thus I used the Generalized Linear Model for this and got significant results both for each factor alone and for the interaction between them, which I interpret as that within the habitats, there is a difference, i.e. dolphins seem to "prefer" some more than others. The same for locals. The significant result of the interaction I interpret as though habitats can be important for dolphins but on specific locals. From this, what I would like to know is which habitats and locals are important, thus I need some post-hoc test similar to Tukey but for non-parametric data, otherwise I don't know what else I can do with the data. That's where I need your help! What post-hoc can I use for what I need on SPSS? Would be very appreciated!
I your situation I would use simple U Mann-Whitney correcting for multiple comparisons like in Bonferroni or Scheffe. It's a hand work, but you could build simple spreadsheet tool for it in excel. No a rocket science.