[R-sig-eco] LRT tests in lmer

[Chris Mcowen]

Rob,

I am having a similar problem.

I posted a question on another R- list and got this response

> Dear Chris,
> 
> Over-dispersion does not occur with a binary response variable so you don't need to test for it.
> 
> This does not mean that between-datum heterogeneity in the probability of success is absent, only that it cannot be observed. For example, take 1000 random draws from a binomial distribution with constant probability (0.5):
> 
> table(rbinom(1000, 1, 0.5))
> 
> and compare the frequency of outcomes with a 1000 draws from 1000 binomial distributions with different probabilities of success (but with mean = 0.5)
> 
> table(rbinom(1000, 1, runif(1000)))
> 
> The data look the same, and so the between-datum heterogeneity (residual variance if you like) although it may exist cannot be estimated from the data.
> 
> Cheers,
> 
> Jarrod

I have taken it that i do not need to test for over-dispersion as i have heard nothing further. However, i am unsure how you test for goodness of fit with a lmer model, i am looking to do it to - so i can see how well my model can be used to predict a response. 

Chris
On 10 Aug 2010, at 15:37, Rob Knell wrote:

Hello all

I apologise for asking what is probably a stupid question, but I've been going round and round in circles for a while now trying to work this out. I'm fitting a model to some data that unfortunately require a GLMM: repeated measures data with binomial errors. I'm a long way out of my comfort zone with this sort of analysis but fortunately the sample size is nice and big and the effects seem pretty clear. I need to do a couple of hypothesis tests to determine what my final model should be (I'm aware of the debate over model simplifaction and so on, let's just say that in this case I'm pretty sure that you'd end up with the same answer whatever you did and if I use a hypothesis testing approach rather than, say, AIC-based model selection I'll at least have consistency with some other analysis). Reading around I find that I need to test for overdispersion but I can't find a clear description of how to do this - can anyone enlighten me? Assuming there isn't a problem with overdi!
spersion, am I OK to use the LRT that anova() gives when I compare a pair of models to test ?

Thanks for any help, I apologise for my numpty-ness but I need to get this done soon and it's not exactly the easiest thing to work out.

Rob


School of Biological and Chemical Sciences
Queen Mary, University of London

'Phone +44 (0)20 7882 7720
Skype Rob Knell

Research: http://webspace.qmul.ac.uk/rknell/

_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology