How do I correctly analyze my live, dead, missing individuals between sites?

Question

I have several study sites, 4 from 2009, 2 from 2010, and 5 from 2011, for a total of 11 sites. At each site, I have 3 quadrats (defined unit of space), and within each quadrat I placed 25 live oysters (75 then at each site). After 1 year, I went back to see how many I could find alive and dead, and then was able to calculate how many went missing (washed away by waves, animal ate, person took, etc). So what I do not know, is what the proper test is to run to compare between the different sites.

Someone once told me to do a contingency table analysis, another told me chi square, and also regression. If you have any suggestions or help that would be great. Thanks, Brian

Answer 1

Since you have counts that are presumably accurate for alive, dead and missing at each site you can construct a 3x3 contingency table for your analysis. The columns would represent each of the three sites and the rows would be the categories alive, dead and missing. Tests like Fisher's exact test or the approximate chi-square test would be options to use to determine if the distributions of alive, dead and mssing differ among sites. I will not go into the details or the assumptions for each test. A really nice example is given by Zen's answer to this recent question which also includes some informative discussion.

The recommendation for contingency table analysis and the one that suggested chi-square were probably both really the same as the approach I have described. Exactly how regression might be applied I am not sure (unless the person is thinking of multinomial logistic regression with site as a covariate).