[R-sig-ME] PedigreeMM question

[David Duffy]

On Thu, 1 Mar 2012, Joehanes, Roby (NIH/NHLBI) [F] wrote:

> Hi,
>
> I would like to do something like this in pedigreemm:
>
> To my impression after reading pedigreemm's code, I need several 
> observations with the same ID number and I cannot have each observation 
> to have a different ID. Is this correct?
>
> What I am getting at is genome-wide association study (GWAS)-like 
> analysis that account for familial correlation matrix. Here, each 
> observation comes from one individual, which is associated an individual 
> ID and a pedigree ID. One pedigree ID
>
> The questions are: How can I deal with this problem? Should I break up 
> the pedigree into nuclear families? How can I deal with the step-parents 
> and intermarriage complications above? If pedigreemm was not designed 
> for problems of this class, which other packages should I look into?
>

The pedigree() constructor is set up as animal breeders do things: the 
entire dataset is one big pedigree, even though there may be subcomponents 
that do not connect to one another.  This is why sparse matrix 
representations are used.  So, you just need unique IDs for everybody (eg 
concatenate pedigree with individual ID).  This allows for intermarriage 
etc.  If you think step-parents transmit your phenotype to step-children, 
then you need to add in a family environmental random effect (ie common to 
all members of an extended family, or perhaps just to that "nongenetic" 
nuclear family).  If you have repeat measures on individuals, then you 
include multiple records with the same unique ID.

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