How to modify slots lme4 >1.0

2019-02-28 12:15发布

I have been using the code below to successfully modify the 'Zt', 'L', and 'A' slots of models fit using lme4 versions <1.0. I just updated to lme4 1.0-4 today and found that the model objects are different. Can anyone provide insight/guidance as to how to modify these slots in the new lmer model objects?

dat<-structure(list(pop1 = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 10L, 
                    2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 4L, 
                    4L, 4L, 4L, 4L, 4L, 5L, 5L, 5L, 5L, 5L, 6L, 6L, 6L, 6L, 7L, 7L, 
                    7L, 8L, 8L, 9L), pop2 = c(2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 1L, 
                    3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 
                    5L, 6L, 7L, 8L, 9L, 10L, 6L, 7L, 8L, 9L, 10L, 7L, 8L, 9L, 10L, 
                    8L, 9L, 10L, 9L, 10L, 10L), X = c(0.49136, 0.75587, 0.93952, 
                    0.61278, 0.79934, 1.07918, 1.13354, 1.15836, 1.2014, 0.43136, 
                    0.77815, 0.716, 0.93952, 1.13672, 1.16137, 1.18184, 1.21748, 
                    0.65321, 0.86332, 1.04922, 1.19866, 1.20412, 1.22272, 1.24797, 
                    0.89763, 1.08991, 1.19033, 1.15836, 1.17319, 1.18752, 0.64345, 
                    0.93952, 0.98227, 1.01703, 1.07188, 0.78533, 0.94939, 0.99564, 
                    1.06819, 0.64345, 0.716, 0.85126, -0.04576, 0.4624, 0.30103), 
           Y = c(0.491694, 0.394703, 0.113303, 0.156597, 0.450924, 0.487845, 
                 0.21821, 0.129027, -0.131522, 0.35156, -0.116826, 0.18941, 
                 0.306608, 0.258401, 0.008552, -0.024369, -0.305258, -0.013628, 
                 0.215715, 0.13783, 0.467272, 0.088882, 0.084295, -0.172337, 
                 -0.206725, -0.084339, -0.191651, -0.001586, -0.079501, -0.195094, 
                 0.232045, 0.17102, 0.003742, -0.023688, -0.26085, 0.205326, 
                 0.172809, 0.133219, -0.159054, 0.082231, 0.011025, -0.238611, 
                 0.732679, 0.478058, 0.325698)), .Names = c("pop1", "pop2", 
                  "X", "Y"), class = "data.frame", row.names = c(NA, -45L))

library(lme4) # lme4 versions >1.0 have different model output    

# Specify the model formula 
    lmer_mod <- as.formula("Y ~ X + (1|pop1)") 

# Create the Zl and ZZ matrices 
    Zl <- lapply(c("pop1","pop2"), function(nm) Matrix:::fac2sparse(dat[[nm]], "d", drop=FALSE)) 
    ZZ <- Reduce("+", Zl[-1], Zl[[1]]) 

# Fit lmer model to the data 
    mod <- lmer(lmer_mod, data = dat, REML = TRUE) 

# Replace the following slots in the fitted model
# These slots don't exist in this form in the new lmerMod objects
    mod@Zt <- ZZ 
    mod@A <- ZZ 
    mod@L <- Cholesky(tcrossprod(ZZ), LDL=FALSE, Imult=1) 

# Refit the model to the same response data 
    Final.mod <- refit(mod, dat[,Y])

Any help or insight as to how to modify these slots will be greatly appreciated. In the meantime, I guess I will stick to using an older version of lme4 for these models.

标签: r lme4 lmer
1条回答
你好瞎i
2楼-- · 2019-02-28 12:38

Does this do what you want? (This follows ?modular pretty closely ...)

Create the Zl and ZZ matrices:

Zl <- lapply(c("pop1","pop2"),
         function(nm) Matrix:::fac2sparse(dat[[nm]], "d", drop=FALSE)) 
ZZ <- Reduce("+", Zl[-1], Zl[[1]]) 

Construct the random-effects data structures:

lf <- lFormula(Y ~ X + (1|pop1), data=dat)

Modify them:

lf$reTrms$Zt <- ZZ

Proceed through the remaining model-construction and fitting steps:

dfun <- do.call(mkLmerDevfun,lf)   ## create dev fun from modified lf
opt <- optimizeLmer(dfun)          ## fit the model
## make the results into a 'merMod' object
fit <- mkMerMod(environment(dfun), opt, lf$reTrms,
         fr = lf$fr)
查看更多
登录 后发表回答