[R] how to do association study based on mixed linear model

[Ben Bolker]

Yao He <yao.h.1988 <at> gmail.com> writes:

> 
> Dear All:
> 
> I want to do association study based on mixed linear model,
> 
> My model not only includes serval fixed effects and random effects but
> also incorporates some covariates such as "birth weight".
> Otherwise, the size of the data are about 180 individuals and 12
> variables and 60000 Fixed effect estimates
> 
> As asreml-R is not free ,is there any packages for my study?
> I heard  nlme or lme4 but I'm not sure whether they could incorporate
> covariates and what about their computational efficiency?

  I believe ASREML-R is free *for noncommercial use* (your address
is at a university, so you might qualify).

  It looks like Windows only, and you need a valid university
e-mail:

http://www.vsni.co.uk/software/free-to-use/teaching/asreml-teaching

  If that doesn't work, then lme4 might (although are you sure
you need 60K *fixed* effects?  Sounds like you need some kind
of penalized regression approach -- or maybe you're going to
do massive multiple comparisons correction??)  You should definitely
follow Bert Gunter's advice in this case and read some documentation
before asking a more focused question on the r-sig-mixed-models
list