In this section we will analyze data from a fictitious experiment
that measured visual contrast sensitivity in six male and six female
children at 30, 36, 42, and 48 months of age. Each child was tested at
each age in two conditions that differed in brightness. The
between-subjects factor (gender
) and the two within-subject
factors (age
and brightness
) are fixed,
whereas subjects is random. The data are in the file
visDevelopment.rda
, which contains the data frames
visDevWide
and visDev
. The data frame
visDev
also contains a numeric variable, csf
,
which contains the dependent variable.
options(contrasts=c("contr.sum","contr.poly"))
load(url("http://pnb.mcmaster.ca/bennett/psy710/datasets/visDevelopment.rda") )
summary(visDev)
## subj gender age brightness csf
## s1 : 8 female:48 a30:24 high:48 Min. : 68.55
## s2 : 8 male :48 a36:24 low :48 1st Qu.: 91.08
## s3 : 8 a42:24 Median :102.34
## s4 : 8 a48:24 Mean :102.49
## s5 : 8 3rd Qu.:113.00
## s6 : 8 Max. :133.00
## (Other):48
sapply(visDevWide,class)
## subj gender high.30 high.36 high.40 high.48 low.30 low.36 low.42 low.48
## "factor" "factor" "integer" "integer" "integer" "integer" "numeric" "numeric" "numeric" "numeric"
gender
,
age
and brightness
on csf
.
Conduct two analyses: a split-plot analysis that includes the
between-subjects variable and a within-subjects analysis that does not.
Where necessary, correct the \(p\)
values for deviations from sphericity. Do all of the \(p\) have to be adjusted? Why or why
not?Use lmer
in the lmerTest
package to
evaluate the effects of gender
, age
, and
brightness
(and interactions) with a mixed model. Treat
subj
as a random factor and all the other factors as fixed.
Evaluate the fixed effects with \(F\)
tests that assume sphericity and chi-square
tests that do
not assume spericity.
Check the residuals of your split-plot model to see if they are distributed normally
Estimate association strength and effect size for each fixed effect in the split-plot model.
Evaluate the linear trend of csf
across age. Is the
overall linear trend (ignoring brightness and gender) significant? Does
the linear trend differ between genders? Does it differ between
brightness conditions?