Hello,
I have little experience with mixed-effects models and could do with some help
when it comes to checking model assumptions.
I am using the 'lme()' function from the 'nlme' library in R to test the fixed
effects of a repeated measures design. Given a pair of sounds (A and B), each
subject was asked to match the loudness of both sounds by adjusting the amplitude of the variable sound.
A number of different pairs were tested, with both A and B as the
variable, and each subject was tested twice in every condition. There are a
total of 160 observations per subject, and 14 participants in all.
Dependent variable: amplitudeDifference
Independent variables: pairOfSounds, variable
The model:
amplitudeDifference ~ pairOfSounds + variable + interaction + (1|Subject/pairOfSounds/variable)
where '(1|Subject/pairOfSounds/variable)' allows for random intercepts of the
factors within each subject.
I have built the model up by AIC/BIC and likelihood ratio tests (method =
"ML"). Residual plots for the within-group errors can be found here:
goo.gl/jfTeKG
The top left plot shows standardised residuals vs fitted values. I don't see any
systematic pattern in the residuals, so I assume that the constant variation assumption
is valid, although further inspection of the subject-by-subject residuals do
show some unevenness. I did try to apply weights to model this but lme() failed to
converge. In conjunction with the top right plot, I have no reason to suspect
non-linearities.
My main concern is the lower left qqplot which reveals that the residuals are
heavy-tailed. I'm not sure where to go from here. From reading Pinheiro and Bates
(2000, p. 180), the
fixed-effects tests tend to be more conservative when the tails are
symmetrically distributed. So perhaps I'm OK if the p-values are very low?
The level two and three random effects show a similar departure from normality.
Any help is much appreciated and thanks for your time in reading this.
I have little experience with mixed-effects models and could do with some help
when it comes to checking model assumptions.
I am using the 'lme()' function from the 'nlme' library in R to test the fixed
effects of a repeated measures design. Given a pair of sounds (A and B), each
subject was asked to match the loudness of both sounds by adjusting the amplitude of the variable sound.
A number of different pairs were tested, with both A and B as the
variable, and each subject was tested twice in every condition. There are a
total of 160 observations per subject, and 14 participants in all.
Dependent variable: amplitudeDifference
Independent variables: pairOfSounds, variable
The model:
amplitudeDifference ~ pairOfSounds + variable + interaction + (1|Subject/pairOfSounds/variable)
where '(1|Subject/pairOfSounds/variable)' allows for random intercepts of the
factors within each subject.
I have built the model up by AIC/BIC and likelihood ratio tests (method =
"ML"). Residual plots for the within-group errors can be found here:
goo.gl/jfTeKG
The top left plot shows standardised residuals vs fitted values. I don't see any
systematic pattern in the residuals, so I assume that the constant variation assumption
is valid, although further inspection of the subject-by-subject residuals do
show some unevenness. I did try to apply weights to model this but lme() failed to
converge. In conjunction with the top right plot, I have no reason to suspect
non-linearities.
My main concern is the lower left qqplot which reveals that the residuals are
heavy-tailed. I'm not sure where to go from here. From reading Pinheiro and Bates
(2000, p. 180), the
fixed-effects tests tend to be more conservative when the tails are
symmetrically distributed. So perhaps I'm OK if the p-values are very low?
The level two and three random effects show a similar departure from normality.
Any help is much appreciated and thanks for your time in reading this.