[R-sig-ME] {nlme} Question about modeling Level two heteroscedasticity in HLM

[Phillip Alday]

I'm not sure, but I think

1. the random slopes may be able to handle this without really doing
anything special

2. in a variation on that idea, you could also the same variable as a
fixed effect and as a grouping variable for a random intercept (see e.g.
Thierry Onkelinx's quadratic example:
http://rpubs.com/INBOstats/both_fixed_random )

3. in yet another variation on that same idea, you could also have an
observation-level random effect.

There may additionally be ways of modelling this by specifying some
covariance structure in nlme.

Phillip

On 08/17/2017 03:53 PM, b88207001 at ntu.edu.tw wrote:
> 
> 
> Hello dear uesRs in mixed models,
> 
> I am working on modeling both level one and level two heteroscedasticity
> in HLM. In my model, both error variance and variance of random
> intercept / random slope are affected by some level two variables.
> 
> I found that nlme is able to model heteroscedasticity. I learned how to
> use it for level one heteroscedasticity but don't know how to use it to
> model the level two heteroscedasticity. I wonder if anyone could give me
> some guide how to do it or any suggested software / packages (such as
> lme4?) if nlme is unable to do it.
> 
> Many thanks in advance!
> 
> Best,
> Yen
> 
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models