doBy (version 4.1.2)

lsmeans: Calculates LSMEANS (population means)

Description

LSMEANS is SAS-jargon for what is sometimes called population means.

Usage

lsmeans(object, eff, at = NULL)
## S3 method for class 'lm':
lsmeans(object, eff, at = NULL)

Arguments

object

A model object; currently only available for 'lm' objects

eff

A vector of variables for which the LSMEANS should be calculated.

at

A list of covariate values

Value

• An 'lsmeansTable' object which is a dataframe with a few additional attributes.

Warning

This function is a recent addition to the 'doBy' package and it is not thoroughly tested. Use with caution!

Details

FIXME: To be written

See Also

FIXME: To be added

Examples

dat <- expand.grid(list(AA=factor(1:2), BB=factor(1:3), CC=factor(1:3)))
dat$y <- rnorm(nrow(dat))
dat$x <- rnorm(nrow(dat))
dat$x2 <- dat$x^2

## Examples

## 1) LSMEANS with factors only.
mod1 <- lm(y ~ AA + BB*CC + x, data=dat)
## Average over AA for each combination of (BB,CC); evaluate at x=mean(x)
lsmeans(mod1, c("BB","CC"))
## Average over (AA,BB) for each value of CC; evaluate at x=mean(x)
lsmeans(mod1, c("CC"))

## 2) The call to lsmeans() below is eqivalent to the following SAS code
## proc glm data=dat;
##	class AA BB CC;
##	model y = AA BB|CC x x*x;
##  lsmeans CC BB*CC / stderr;
## run;
##
mod2 <- lm(y ~ AA + BB*CC + x + x2, data=dat)
lsmeans(mod2, "CC")

## Notice the difference to:
mod3 <- lm(y ~ AA + BB*CC + x + I(x^2), data=dat)
lsmeans(mod3, "CC")

## The difference arises because in the former case, x2 is evaluated at mean(x2) whereas
## in the latter case x is evaluated at mean(x)^2

## 3) Plug in particular values of covariates
## The call to lsmeans() below is eqivalent to the following SAS code
## proc glm data=dat;
##	class AA BB CC;
##	model y = AA BB|CC x x2;
##  lsmeans CC BB*CC / at x=2 stderr;
## run;
lsmeans(mod2, c("CC"), at=list(x=2))
## Above, x=2 is used while x2 is set to mean(x2)

## The call to lsmeans() below is eqivalent to the following SAS code
## proc glm data=dat;
##	class AA BB CC;
##	model y = AA BB|CC x x*x;
##    lsmeans CC BB*CC / at x=2 stderr;
## run;
##
lsmeans(mod3, c("CC"), at=list(x=2))
## Above, x=2 is used while I(x^2) is set to mean(x^2)=4. Notice that setting
## lsmeans(mod3, c("CC"), at=list(x=2,"I(x^2)"=123))
## has no effect: I(x^2) is still 4 because x=2. Hence the following two results
## are identical

lsmeans(mod2, c("CC"), at=list(x=2, x2=4))
lsmeans(mod3, c("CC"), at=list(x=2))

## END