0
$\begingroup$

I contact you because my case is particular and I don’t know much about GLMM. I have data of social networks (network metrics) of a nonhuman primate species. These data are by nature non independent (behaviour between A and B can influence those of C). I also have individual attributes (gender, age, kinship and hierarchical rank). I would like to test the effect of individual attributes on the network metrics. For this purpose I have to take into count the non-independency of data (I could make link filtering or permutation test but I am trying something new). I was thinking about doing a General Linear Mixed Model (GLMM) including individuals as random factors. I see that this analysis can take into count the non-independency of data, but I only see this test for multiple measures on same individuals (which is another kind of non-independency) and not for a single measure on individuals. Do you think even if this is not the same type of non-independency data GLMM can take into count my type of data?

Regards


Thank you for your answer. I watched p model but it seems that this method needs binary networks. I also know MRQAP or QAP analysis but as for p model they correlate networks with indiviudal attributes and not network metrics with attributes. I also know anova and t-test with permutation performed on ucinet but I would like to use a new protocol and I think GLMM can make up with independent data. This idea emerges from Darren Croft's paper 'Hypothesis testing in animal social networks' in which he says: 'Model fitting is a potentially different approach to hypothesis testing. In particular, mixed models can account for some forms of data non-independence and have already been applied to the analyses of animal social networks'.

Regards

$\endgroup$
0
$\begingroup$

I would recommend using either p*/Exponential Random Graph Modeling or Quadratic Assignment Procedure tests over GLMM. Those methods are made especially for the data you have available.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.