[R-sig-ME] Replicating analysis in asreml in lme4

[Ben Bolker]

 Try the lsmeans package ...

On 16-07-30 01:41 AM, Aravind J. wrote:
> Dear members,
> 
> I am trying to reproduce analyse an alpha lattice design (an unbalanced
> design) in  `lme4`.
> 
> https://cran.r-project.org/web/packages/agridat/agridat.pdf#page.184
> 
> library(agridat)
> library(lme4)
> 
> The model can be fitted in lme4
> 
> # genotypes - fixed
> model <- lmer(yield ~ 0 + gen + rep + (1|rep:block), dat)
> # genotypes - random
> model <- lmer(yield ~ 0 + (1|gen) + rep + (1|rep:block), dat)
> 
> However further calculations seem to required `asreml`.
> 
> library(asreml)
> 
> m3 <- asreml(yield ~ 1 + rep, data=dat, random=~ gen + rep:block)
> sg2 <- summary(m3)$varcomp[1,'component']
> vblup <- predict(m3, classify="gen")$pred$avsed ^ 2
> 
> m3 <- asreml(yield ~ 1 + gen + rep, data=dat, random = ~ rep:block)
> vblue <- predict(m3, classify="gen")$pred$avsed ^ 2
> sg2 / (sg2 + vblue/2)
> 1-(vblup / 2 / sg2)
> 
> Here `avsed` is the "mean variance of difference of adjusted means" (BLUP
> or BLUE). Is it possible to replicate the last part also in `lme4`?
> 
> `avsed` can be computed from the "variance-covariance matrix of adjusted
> means" for genotypes. (
> https://static-content.springer.com/esm/art%3A10.1186%2F1471-2164-14-860/MediaObjects/12864_2013_5591_MOESM1_ESM.doc).
> How to get that matrix from lme4?
> 
> 
> Warm Regards,
>