lmer_bs function

Sampling of bootstrap data from a given random effects model

lmer_bs() draws bootstrap samples based on the estimates for the mean and the variance components drawn from a random effects model fit with lme4::lmer(). Contrary to lme4::bootMer(), the number of observations for each random factor can vary between the original data set and the bootstrapped data. Random effects in model have to be specified as (1|random effect).

lmer_bs(model, newdat = NULL, futmat_list = NULL, nboot)

Arguments

• model: a random effects model of class lmerMod
• newdat: a data.frame with the same column names as the historical data on which model depends
• futmat_list: a list that contains design matrices for each random factor
• nboot: number of bootstrap samples

Returns

A list of length nboot containing the bootstrapped observations.

Details

The data sampling is based on a list of design matrices (one for each random factor) that can be obtained if newdat and the model formula are provided to lme4::lFormula(). Hence, each random factor that is part of the initial model must have at least two replicates in newdat.

If a random factor in the future data set does not have any replicate, a list that contains design matrices (one for each random factor) can be provided via futmat_list.

Examples

# loading lme4
library(lme4)

# Fitting a random effects model based on c2_dat1

fit <- lmer(y_ijk~(1|a)+(1|b)+(1|a:b), c2_dat1)
summary(fit)

#----------------------------------------------------------------------------

### Using c2_dat2 as newdat

c2_dat2

lmer_bs(model=fit, newdat=c2_dat2, nboot=100)

#----------------------------------------------------------------------------

### Using futmat_list

# c2_dat4 has no replication for b. Hence the list of design matrices can not be
# generated by lme4::lFormula() and have to be provided by hand via futmat_list.

c2_dat4

# Build a list containing the design matrices

fml <- vector(length=4, "list")

names(fml) <- c("a:b", "b", "a", "Residual")

fml[["a:b"]] <- matrix(nrow=6, ncol=2, data=c(1,1,0,0,0,0, 0,0,1,1,1,1))

fml[["b"]] <- matrix(nrow=6, ncol=1, data=c(1,1,1,1,1,1))

fml[["a"]] <- matrix(nrow=6, ncol=2, data=c(1,1,0,0,0,0, 0,0,1,1,1,1))

fml[["Residual"]] <- diag(6)

fml

lmer_bs(model=fit, futmat_list=fml, nboot=100)