|
|
Hello!
I have a question regarding model comparison. I have created a habitat suitability index based on a number of variables that characterize a landscape on a scale from 1-5. I have added them together in different ways to get a series of models (eg. one is composed of variables a and b, one of a, b, c; one is a,b,d and so on). I want to test whether or not the "presence" locations I have for my critter of interest are more likely to be classified as 5's to verify that if the model worked/accurately predicted presence, and to see which model best predicted it. I am using a chi-square test, although I probably need to use a Fisher's Exact Test due to low sample sizes (unfortunately I do not have the Exact Test module, however). But that is another issue. I am wondering if I can simply run this statistic for every possible model and then compare the p values to see which gives the most significant deviation from expected values. It seems a little too straightforward to actually work that way, but it makes intuitive sense to me. I appreciate any help!
Thanks, Amanda
|
Locked
