AICcmodavg (version 1.15)

predictSE.mer: Computing Predicted Values and Standard Errors

Description

Function to compute predicted values based on fixed effects (population predictions) and associated standard errors from a generalized linear mixed effect model.

Usage

predictSE.mer(mod, newdata, se.fit = TRUE, type = "response", level = 0, 
print.matrix = FALSE)

Arguments

mod

an object of class 'mer' containing the output of a model.

newdata

a data frame with the same structure as that of the original data frame for which we want to make predictions.

se.fit

logical. If TRUE, compute standard errors on predictions.

type

specifies the type of prediction requested. This argument can take the value 'response' or 'link', for predictions on the scale of the response variable or on the scale of the linear predictor, respectively.

level

the level for which predicted values and standard errors are to be computed. The current version of the function only supports predictions for the populations excluding random effects (i.e., level = 0).

print.matrix

logical. If TRUE, the output is returned as a matrix, with predicted values and standard errors in columns. If FALSE, the output is returned as a list.

Value

• 'predictSE.mer' returns requested values either as a matrix ('print.matrix = TRUE') or list ('print.matrix = FALSE') with components:
• fit
  the predicted values.
• se.fit
  the standard errors of the predicted values (if 'se.fit = TRUE').

Details

'predictSE.mer' computes predicted values based on fixed effects and associated standard errors. Standard errors are approximated using the delta method (Oehlert 1992).

References

Gelman, A., Hill, J. (2007) Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press: New York. Oehlert, G. W. (1992) A note on the delta method. American Statistician 46, 27--29.

See Also

predictSE.lme, mer-class, lmer, glmer

Examples

##contagious bovine pleuropneumonia example modified from lme4
require(lme4)
data(cbpp)
##create proportion of incidence
cbpp$prop <- cbpp$incidence/cbpp$size
gm1 <- glmer(prop ~ period + (1 | herd), family = binomial, weights =
size, data = cbpp)

##create a data set to make predictions
newherd<- data.frame(period = as.factor(c("1", "2", "3", "4")))

##predictions on logit link scale
predictSE.mer(mod = gm1, newdata = newherd, se.fit = TRUE, type =
"link", level = 0, print.matrix = FALSE)

##predictions on scale of original response variable
predictSE.mer(mod = gm1, newdata = newherd, se.fit = TRUE, type =
"response", level = 0, print.matrix = TRUE)


##example with linear mixed model with Orthodont data from
##Pinheiro and Bates (2000)
data(Orthodont, package = "nlme")
m2 <- lmer(distance ~ Sex + (1 | Subject), data = Orthodont, REML =
FALSE)

##create data set for predictions for all combinations of Sex and 2 ages
neworth <- expand.grid(Sex = c("Male", "Female"), age = c(8, 10))

##compute predicted values
predictSE.mer(m2, newdata = neworth)
##the following yields the same answer because the model
##uses an identity link
predictSE.mer(m2, newdata = neworth, type = "link") 

predictSE.mer(m2, newdata = neworth, level = 1, print.matrix = TRUE)
##generates an error

##compare with the following:
m3 <- glmer(distance ~ Sex + (1 | Subject), family = gaussian(link =
log), data = Orthodont, nAGQ = 1)
##predictions on original scale of response variable
predictSE.mer(m3, newdata = neworth, type = "response")
##predictions on log scale
predictSE.mer(m3, newdata = neworth, type = "link")

detach(package:lme4)