Skip to content
Prev 19070 / 20628 Next

MCMCglmm covariance matrix specification

Hi,

As I understand it you have 3 2x2 covariance matrices to be estimated,
one for each environment?

~us(at.level(env, 1):trait):animal+us(at.level(env,
2):trait):animal+us(at.level(env, 3):trait):animal

should work. I presume you have no shared pedigree between the
envrionments hence the cross-env covariances are not estimable? In the
computation time is long, get back to me; there are ways to
reparameterise it to make it faster but it's a bit fiddly.

For harder problems (where the covariance matrix can't be permuted such
that the estimable bits fall in blocks along the diagonal, as here) then
fixing elements to zero is probably not a good idea even if you could do
it (for example in asreml). The zero elements will force patterns in the
estimable elements to ensure positive-defitness. The antedependence
solution I posted earlier gets round this issue I believe.

Cheers,

Jarrod
On 24/02/2021 18:29, Walid Crampton-Mawass wrote:
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. Is e buidheann carthannais a th? ann an Oilthigh Dh?n ?ideann, cl?raichte an Alba, ?ireamh cl?raidh SC005336.