Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 1 down vote favorite

Is it possible to do discriminant analysis with random effects? Is there an R package for this?

Context:

I have habitat use data for two species of frogs from radio telemetry, but nested within 'species' the data is for individuals with highly autocorrelated data.

For example, I have ~280 relocations for species 1, with a mean of 20 relocations per individual, and ~ 210 relocations for species 2, with a mean of 13 relocations per animal. Relocations per animal are highly autocorrelated. At each relocation I collected structural habitat data, and I would like to identify which parameters are most strongly identified with each species.

My preference is to have this autocorrelation accounted for in the analysis, but my alternatives with DA, I think, are to:

  1. Use mean values for each individual, however then I would be breaking assumptions of sample size as I would have fewer samples than variables (therefore not a viable alternative); or
  2. Cut relocations from individuals with lots of data points.

I've done some searching through CranR, docs, etc, but haven't had much success. The MASS package does discriminant analysis and random effects, but RE in DA? I'd be grateful to be pointed in the right direction (or told that I'm completely out to lunch!)

share|improve this question
Update: I haven't found any evidence that this is possible, so I've taken the means for each individual animal and will use that as my data matrix. – Monica Nov 13 '12 at 23:16
Update: I haven't found any evidence that this is possible, so I've taken the means for each individual animal and will use that as my data matrix and slashed the variables I'm using. Unfortunately, the data are non-normal and variances are not homogeneous across groups, for biological reasons. I've seen that there are workarounds for heterogeneity (hda / qda), and that qda may also work on non-normal data. I'm wondering if I'm going too far down the statistical rabbit hole with all these workarounds and if there's another obvious method that I'm missing.. – Monica Nov 13 '12 at 23:23
Update: I've moved on from DA, as it seems GLMMs are more suited to my analysis, but I'm coming up against some issues with that as well. Questions regarding the GLMM challenges will be in a different thread. Thanks for reading.. – Monica Nov 30 '12 at 3:08

Know someone who can answer? Share a link to this question via email, Google+, Twitter, or Facebook.

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.