R: Calculate population means (LSMEANS in SAS jargon)

popMeans {doBy}

R Documentation

Calculate population means (LSMEANS in SAS jargon)

Description

Calculate population means (LSMEANS in SAS jargon).

Usage

popMeans(object, effect, at = NULL, only.at=TRUE, engine=esticon,grid=TRUE,...)
## Default S3 method:
popMeans(object, effect, at = NULL, only.at=TRUE, engine=esticon,grid=TRUE,...)
## S3 method for class 'lme'
popMeans(object, effect, at = NULL, only.at=TRUE, engine=esticon,grid=TRUE,...)

Arguments

`object`	A model object; currently only available for 'lm' objects
`effect`	A vector of variables for which the population means should be calculated.
`at`	A list of covariate values
`only.at`	If FALSE and `at` is not NULL then the contrast matrix is formed as if `at` was not specified. Then the values of `at` are set afterwards such that the contrast matrix will contain several identical rows.
`engine`	The function for actually calculating the linear combination of the parameters. See 'details'.
`grid`	If `TRUE` and `engine='esticon'` then the values of the covariates at which the population means are calculated are appended to the result.
`...`	Additional arguments; currently not used.

Details

The function is lsMeans() is just an alias for popMeans().

The popMeans() function generates the relevant contrast matrix and then calls the engine on that matrix. Another possible engine is the glht function from the multcomp package

Value

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!

Author(s)

Søren Højsgaard, sorenh at mail dot dk

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)
popMeans(mod1, c("BB","CC"))
## Average over (AA,BB) for each value of CC; evaluate at x=mean(x)
popMeans(mod1, c("CC"))

## 2) The call to popMeans() below is equivalent 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)
popMeans(mod2, "CC")

## Notice the difference to:
mod3 <- lm(y ~ AA + BB*CC + x + I(x^2), data=dat)
popMeans(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 popMeans() below is equivalent 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;
popMeans(mod2, c("CC"), at=list(x=2))
## Above, x=2 is used while x2 is set to mean(x2)

## The call to popMeans() below is equivalent 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;
##
popMeans(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
## popMeans(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

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

## END

[Package doBy version 4.5.2 Index]