itsadug (version 2.4.1)

wald_gam: Function for post-hoc comparison of the contrasts in a single GAMM model.

Description

Function for post-hoc comparison of the intercept differences for different factors in a single GAMM model.

Usage

wald_gam(
  model,
  comp = NULL,
  select = NULL,
  t.test = FALSE,
  null.hypothesis = 0,
  summ = NULL,
  signif.stars = TRUE,
  print.output = getOption("itsadug_print")
)

Arguments

model

Model, currently only implemented for models generated with bam or gam.

comp

Named list with predictors (specified as names) and their levels to compare. Defaults to NULL, which returns all comparisons, unless select is specified.

select

Contrast matrix for manually specified contrasts. Alternatively, a vector or list could be provided as input. See examples below.

t.test

Logical default = FALSE), whether or not to return the t-test scores instead of the Wald test. Only implemented for Gaussian models. This option is not implemented for use with select.

null.hypothesis

Numeric, value of null hypothesis. Defaults to 0 and is generally not changed.

summ

Optional summary object. Defaults to NULL. For very large GAMM models it takes a long time to retrieve the summary. In these cases the summary could be provided to reduce processing time. However, it is generally recommended not to specifify a summary object, to reduce the chance of mismatch errors.

signif.stars

Logical (default = TRUE). Whether or not to display stars indicating the level of significance on 95% confidence level.

print.output

Logical: whether or not to print the output. By default controlled globally in the package options: If the function infoMessages is set to TRUE, the output will be automatically printed. Could be also set by explicitly providing TRUE or FALSE. See examples.

Value

Optionally returns a data frame with test statistics.

Warning

This function is intended for testing intercept differences only. This function compares purely the parametric components, without considering any interactions with smooth terms. So this could be considered as a partial effect comparison. For comparing the averages of conditions use get_difference, which outputs the difference in summed effects for different factor levels.

See Also

plot_parametric, plot_diff, plot_diff2

Other Testing for significance: compareML(), plot_diff2(), plot_diff(), report_stats()

Examples

# NOT RUN {
data(simdat)
# Convert Condition to factorial predictor for illustration purposes:
simdat$Condition <- as.factor(simdat$Condition)

infoMessages('on')

# }
# NOT RUN {
# some arbitrary model:
m1 <- bam(Y ~ Condition*Group  
\t+ s(Time, by=Condition) 
\t+ s(Time, by=Group)
\t+ s(Subject, bs='re'), 
\tdata=simdat)

# print summary to inspect parametric terms:
summary(m1)

# return all contrasts:
wald_gam(m1)

# USE OF COMP
# return only contrasts for Adults:
wald_gam(m1, comp=list(Condition=levels(simdat$Condition)))
# return specific contrasts:
wald_gam(m1, comp=list(Condition=c('-1', '0', '1'), 
    Group=c('Adults', 'Children')))

# USE OF SELECT
# Specify contrast matrix. 
# Note that intercept should be 0.
# Example: Compare Condition 0 with Conditions 2 and 3 for children.
# Method 1: matrix or vector:
R = matrix( c(0,-2,0,1,1,0,0,0,0,0,0,0), nrow=1)
wald_gam(m1, select=R) 
wald_gam(m1, select=c(0,-2,0,1,1,0,0,0,0,0,0,0)) 
# Method 2: list
# first list element are reference coefficients, 
# second list element are coefficients to compare
wald_gam(m1, select=list(2, c(4,5))) 
# Replication of contrasts given in summary:
wald_gam(m1, select=c(0,1,0,0,0,0,0,0,0,0,0,0))

# USE OF T.TEST
# This option is not implemented for use with select
# Compare with second line of parametric summary:
wald_gam(m1, comp=list(Condition=c('-1', '0'), 
    Group='Children'), t.test=TRUE)
# Compare with Wald test:
wald_gam(m1, comp=list(Condition=c('-1', '0'), 
    Group='Children'))

# exclude significance stars:
wald_gam(m1, comp=list(Condition=c('-1', '0'), 
    Group='Children'), signif.stars=FALSE)

# do not print output, but save table for later use:
test <- wald_gam(m1, comp=list(Condition=c('-1', '0'), 
    Group='Children'), print.output=FALSE)
test
# alternative way:
infoMessages('off')
test2 <- wald_gam(m1, comp=list(Condition=c('-1', '0'), 
    Group='Children'))
infoMessages('on')

# }