I have 36 different population types that vary in combinations of three parameters (s, g, d). g and g has 3 discreet variations (0.2, 0.8, 0.95 and 10, 100, 500 respectively) and d has 4 (1.0, 0, 0.1, 1.1). I am doing a power-analysis by analyzing how many times each population type yields an accurate estimate for my factor of interest, out of 500. I would like to do a pair-wise analysis of each population in R to see which populations have a significant difference between them in terms of the accuracy counts... thankfully most of the populations cluster together in 4-5 groups depending on the s parameter and to some degree the g parameter, so I'm really only looking for differences between these, but would still like to do it between all 36 and then output the populations with major differences, which would then allow me to analyze which parameter combinations result in the most accurate results/which parameters have the most influence.
This is a type of table I have.
|Pop||Num of accurate estimates|
etc. for 36 populations. So for pop 1 for example has an s of 0.2, g of 10, and d of 1.0, etc.
I was thinking of doing a chi-square comparison, g-test 1-sample proportion test, which in my understanding would create a pairwise contingency table. But when I research these statistical methods, it seems these are only carried out for two groups. I understand I could just create a piece of code that does a test for each combo and then apply some sort of correction, but I was wondering if any one knows of a specific statistical test that does this, or a direction I could head into?