Re: Order of fitting fixed effects

From: <arthur.gilmour_at_CARGOVALE.COM.AU>
Date: Tue, 28 Jul 2009 18:37:59 +1000

Dear Andrew,

In general, one would test a term adjusting for all other terms not
involving it. So, to test for |SNP, you would adjust for age and HYS (but
in a more complicated model, you would not adjust SNP for a term like
SNP.age).

However, it is sometimes of interest (not necessarily in this case, to
understand when covariates compensate for each other or need each other.

For example you might fit SNP1 and SNP2 in the model. Various things
might happen:

SNP1 | mu NS
SNP2 | mu and SNP1 ***

does not necessarily mean SNP1 is really NS because changing the order
might produce

SNP2 | mu NS
SNP1 | mu SNP2 ***

Two particular points with ASReml.

1) If HYS has more than 800 levels, you would need to specify
the model as
Height ~ mu Age SNP !r ANIMAL !f HYS

Otherwise ASReml would not provide and ┴nalysis of Variance┤table
of Wald F statisitics.

2) The qualifiers !FCON will give tests of fixed effects with maximal
marginality inserting an ezrta column into the ┴NOVA┤table

3) !DDF may be required to get the Denominator DF calculated.

There is debate about these things and I have given my view.
I trust it helps.
> Dear All,
>
>
>
> I am looking for some advice on determining the order in which fixed
> effects should be fitted in an animal model.
>
>
>
> I am currently carrying out association studies between SNPs and growth
> in cows. My fixed effects include herd-year-season (HYS), covariate of
> Age, and the SNP, and a random effect of 'Animal'. The model is
> currently arranged in the following order -
>
>
>
> Height ~ mu HYS Age SNP !r ANIMAL
>
>
>
>
>
> I have found that if I change the order in which the fixed effects are
> fitted, the results differ. After checking various papers, it appears
> that some have fitted the SNP first, whereas others fit the SNP at the
> end.
>
>
>
> I would be very grateful for any clarification,
>
> Kind Regards
>
> Andrew
>
>
>
>
> This message is intended for the addressee named and may contain
> confidential information. If you are not the intended recipient, please
> delete it and notify the sender. Views expressed in this message are those
> of the individual sender, and are not necessarily the views of their
> organisation.
>
Received on Wed Jul 28 2009 - 18:37:59 EST

This webpage is part of the ASReml-l discussion list archives 2004-2010. More information on ASReml can be found at the VSN website. This discussion list is now deprecated - please use the VSN forum for discussion on ASReml. (These online archives were generated using the hypermail package.)