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:
This email was sent to you by someone outside the University. You should only click on links or attachments if you are certain that the email is genuine and the content is safe. Hey all, Hope you are doing well during this time! I have been racking my brain for weeks on how to do model this issue but I have found nothing other than one old answer by Jarrod Hadfield ( https://stat.ethz.ch/pipermail/r-sig-mixed-models/2015q4/024036.html) which recommends using an antedepedence model. Here is the issue: I have constructed a bivariate animal model (trait1, trait2) with a random interaction with the additive genetic random effect and the residual variance,i.e. (trait:env):animal. The interaction variable is a categorical environmental variable of 3 levels (Low, Mid, High). So my variance-covariance matrix has a 6x6 shape (2traitsx3env). Hence, the matrix would include both among-trait covariances within the same env and between env, and cross-env covariances for the same trait: trait1:low trait1:mid trait1:high trait2:low trait2:mid trait2:high 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 (1 represent variances, 0 represent covariances) I have already run the model with both the idh() and us() specification. In the first case, no covariances are calculated at all, only variances are calculated. In the second case, all types of covariances are calculated. I need help figuring out how to specify the variance-covariance matrix in MCMCglmm (and prior) in a way to tell the model not to estimate the cross-env covariances, only the among-trait covariances should be estimated: trait1:low trait1:mid trait1:high trait2:low trait2:mid trait2:high 1 x x 0 x x x 1 x x 0 x x x 1 x x 0 0 x x 1 x x x 0 x x 1 x x x 0 x x 1 (1 represent variances, 0 represent covariances to be estimated, x represent covariances fixed at 0, i.e. not estimated) any help would be appreciated! -- Walid Crampton-Mawass Ph.D. candidate in Evolutionary Biology Population Genetics Laboratory University of Qu?bec at Trois-Rivi?res 3351, boul. des Forges, C.P. 500 Trois-Rivi?res (Qu?bec) G9A 5H7 Telephone: 819-376-5011 poste 3384 [[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
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.