Formatting data for extended family models

Posted on
No user picture. pgseye Joined: 10/13/2009

Hi,

I'm stuck with how one sets up their data frame to take into account relationships beyond that of twins (in my case - sibs and triplets).

For a standard twin model - your data is 'wide' formatted:

ie Var_t1 Var_t2
... ...
... ...

How do you do this for sibs - the closest I can think is to model a sib against each of their twin siblings, something like:

Var_t1 Var_Sib
Var_t2 Var_Sib for one family, then

Var_t1 Var_Sib
Var_t2 Var_Sib for another family, etc

and treat the relationship as you do for a DZ pair.

But then I have some twin pairs with multiple sibs - which would become considerably more complex. And also half a dozen triplets. I'm assuming I'm over-complicating things.

I'd appreciate any help.

Thanks,

Paul

Replied on Tue, 09/07/2010 - 10:04
Picture of user. neale Joined: 07/31/2009

Essentially, the FIML methods assume that data on different rows are independent of each other. Therefore, it is necessary to organize possibly non-independent observations on one line (a wide format). Thus:

Var_t1 Var_t2 Var_Sib1 Var_Sib2

would be an example 'record' or row of the data frame. Note that this means that you need to specify a model with covariance structure capable of describing the largest pedigree, and all smaller ones (which would have missing data coded for non-existent relatives' data).

There is an alternative approach via multilevel modeling, but I don't know of any OpenMx implementation of this for unbalanced pedigrees.