mlmm {LMMstar}

Fit Multiple Linear Mixed Model

Description

Fit several linear mixed models, extract relevant coefficients, and combine them into a single table.

Usage

mlmm(
  ...,
  data,
  by,
  contrast.rbind = NULL,
  effects = NULL,
  robust = FALSE,
  df = TRUE,
  ci = TRUE,
  name.short = c(TRUE, TRUE),
  transform.sigma = NULL,
  transform.k = NULL,
  transform.rho = NULL,
  transform.names = TRUE,
  trace = TRUE
)

Arguments

...

arguments passed to lmm.

data

[data.frame] dataset (in the long format) containing the observations.

by

[character] variable used to split the dataset. On each split a seperate linear mixed model is fit.

contrast.rbind

[character or numeric matrix] Contrast to be be applied to compare the groups. Argument passed to the argument effects of rbind.Wald_lmm. Right hand side can be specified via an attribute "rhs".

effects

[character or numeric matrix] Linear combinations of coefficients relative to which Wald test should be computed. Argument passed to anova.lmm. Right hand side can be specified via an attribute "rhs".

robust

[logical] Should robust standard errors (aka sandwich estimator) be output instead of the model-based standard errors. Argument passed to anova.lmm.

df

[logical] Should the degree of freedom be computed using a Satterthwaite approximation? Argument passed to anova.lmm.

ci

[logical] Should a confidence interval be output for each hypothesis? Argument passed to anova.lmm.

name.short

[logical vector of length 2] use short names for the output coefficients: omit the name of the by variable, omit the regression variable name when the same regression variable is used in all models.

transform.sigma, transform.k, transform.rho, transform.names

[character] transformation used on certain type of parameters.

trace

[interger, >0] Show the progress of the execution of the function.

Details

Grouping variable in argument repetition: when numeric, it will be converted into a factor variable, possibly adding a leading 0 to preserve the ordering. This transformation may cause inconsistency when combining results between different lmm object. This is why the grouping variable should preferably be of type character or factor.

See Also

confint.mlmm for a data.frame containing estimates with their uncertainty.
summary.mlmm for a summary of the model and estimates.
autoplot.Wald_lmm for a graphical display.

Examples

#### univariate regression ####
if(require(lava) && require(multcomp)){

set.seed(10)
d1 <- cbind(sim(lvm(Y~0.5*X1), 25), group = "A")
d2 <- cbind(sim(lvm(Y~0.1*X1), 100), group = "B")
d3 <- cbind(sim(lvm(Y~0.01*X1), 1000), group = "C")
d1$id <- 1:NROW(d1)
d2$id <- 1:NROW(d2)
d3$id <- 1:NROW(d3)

d <- rbind(d1,d2,d3)

e.mlmm <- mlmm(Y~X1, data = d, by = "group", effects = "X1=0")
summary(e.mlmm)
summary(e.mlmm, method = "single-step")
summary(e.mlmm, method = "single-step2")

## re-work contrast
summary(anova(e.mlmm, effects = mcp(X1 = "Dunnett")), method = "none")
## summary(mlmm(Y~X1, data = d, by = "group", effects = mcp(X1="Dunnett")))
}

#### multivariate regression ####
set.seed(10)
dL <- sampleRem(250, n.times = 3, format = "long")

e.mlmm <- mlmm(Y~X1+X2+X6, repetition = ~visit|id, data = dL,
               by = "X4", structure = "CS")
summary(e.mlmm)

e.mlmmX1 <- mlmm(Y~X1+X2+X6, repetition = ~visit|id, data = dL,
               by = "X4", effects = "X1=0", structure = "CS")
summary(e.mlmmX1)
summary(e.mlmmX1, method = "single-step")