[R-sig-ME] Random or Repeated?

[David Duffy]

On Fri, 9 Dec 2011, Kyle Hernandez wrote:

> I am building some mixed models using QTLs for several traits. This 
> basically gets at the underlying genetic architecture.  However, I have 
> a Recombinant Inbred Line (RIL) breeding design grown across 5 
> environments. There are a couple issues I have:
>
> 1. Each RIL is not truly independent from one another (some are more closely related than others)
> 2. Can measuring traits for a RIL across several environments be considered "repeated" measures?

Yes and no ;) I presume you might be interested in QTL x environment 
interactions.  And it sounds like you would like to do a multivariate, ie 
multitrait, analysis.

>
> I originally was using RIL as a random effect in my mixed models, but 
> this kind of removed all the genetic variation from my models. However, 
> perhaps I can fit RIL as a repeated measure with a specified covariance 
> structure? I have also seen a type of "relatedness matrix"  used as a 
> random effect (perhaps some penalized model)?

You might look at the wgaim and mpMap packages that do biparental and 
multiparental RIL analyses respectively.  Use of those packages requires 
you to have ASReml (but you may already have a licence).  Alternatively, 
you can meditate on the papers describing them and reimplement the models 
in lme4...

Cheers, David Duffy.

-- 
| David Duffy (MBBS PhD)                                         ,-_|\
| email: davidD at qimr.edu.au  ph: INT+61+7+3362-0217 fax: -0101  /     *
| Epidemiology Unit, Queensland Institute of Medical Research   \_,-._/
| 300 Herston Rd, Brisbane, Queensland 4029, Australia  GPG 4D0B994A v