Include maternal effect in animal model

[wimgo...@hotmail.com]

Hey

I'm currently using an additive genetic animal model in BreedR to calculate EBVs for vitality in pigs, based on scores of litters. I would like to include a maternal effect, because the mother has a great non-genetic influence on the vitality of their offspring.

I know this is possible via the BLUPF90 software (http://nce.ads.uga.edu/wiki/doku.php?id=readme.renumf90; here you can specify a maternal effect optional to your animal effect).

Is it also possible to include a maternal effect in BreedR?

How would I have to do this?

Kind regards

Wim Gorssen

Junior quantitative geneticist at KU Leuven (Belgium), 

department of Biosystems, research group Livestock Genetics

[Facundo Muñoz]

Dear Wim,

In principle, you could simply use an unstructured random effect of the mother. I.e., if you had a variable "mum" identifying the mother, you would include the argument `random = ~ mum` in your remlf90() call.

However, be careful in interpreting the results as this effect may be confounded with the additive genetic effect. If you have a full and deep pedigree should work. On the other extreme, if you had a partial pedigree with information only about the mother, there will be identifiability issues.

ƒacu.-

--
Report issues at https://github.com/famuvie/breedR/issues
---
You received this message because you are subscribed to the Google Groups "breedR" group.
To unsubscribe from this group and stop receiving emails from it, send an email to breedr+un...@googlegroups.com.
To post to this group, send email to bre...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/breedr/9cb2a342-42b3-4805-a125-576e353c68aa%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Facundo Muñoz]

There is no way to work around unidentifiable models.

As an approximation, you could use the pedigree only, and then summarize the breeding values of the offspring for each mother. The mean of that distribution would be a point estimate of the maternal effect.

For making formal inference on those effects (p-values, CIs, etc.) you would need to use bootstrapping techniques, though.

Hope it helps.

        ƒacu.-

On 03/08/2018 03:39 PM, wimgo...@hotmail.com wrote:

Dear Facundo

I do have a partial pedigree, where the father's side contains a full and deep pedigree, but the mother's side is only known up to 1 generation.

For the moment, I am using the mother as an unstructured random effect, but because of the confounding issues with the additive genetic effect that you mentioned, I was wondering if there was a way in BreedR to work around this.

Op dinsdag 6 maart 2018 17:02:16 UTC+1 schreef Facundo Muñoz:

To view this discussion on the web visit https://groups.google.com/d/msgid/breedr/7682c985-c274-46e0-9ee9-53fc0c89cda4%40googlegroups.com.

[wimgo...@hotmail.com]

Dear Facundo

As an alternative, I am running a sire model with the dam-effect as uncorrelated random effect at the moment.

I was wondering if it would be possible to run a sire-model with a correlated random dam-effect. I've searched the BreedR help site (https://github.com/famuvie/breedR/wiki/Getting-started), but I can't seem to find how to include an extra correlated random effect, unless maybe via the generic component?

Kind Regards

Wim

Op dinsdag 13 maart 2018 16:41:45 UTC+1 schreef Facundo Muñoz:

[Facundo Muñoz]

Yes. If you can compute de correlation matrix for the dams, then you can feed it into a generic model.

Best wishes

        ƒacu.-

To view this discussion on the web visit https://groups.google.com/d/msgid/breedr/1967a3ab-d18b-4b95-ae60-19d027e07d13%40googlegroups.com.