Problem set 10
Complete this task individually or in teams of up to three students.
Submission information: please submit on ZoneCours
- a PDF report
- your code
Read the chocolate example from Meier (2022) and some of the examples from the course notes.
Elliott et al. (2021) attempt to replicate a study of Flavell et al. (1966) and study unprompted verbalization by children aged 5 to 10 in an experiment. Data MULTI21_D2
from package hecedsm
contains the data, including unique participant id
, lab
, age
group (either 5, 6, 7 or 10 year old), different timing
for the recall task (either point-and-name, which is always last, delayed recall with 15 seconds after presenting the last image, or immediate response). You can also download the SPSS database via this link.
We consider the following sources of variation: age
, id
, lab
and timing
.
Fit a linear mixed model for the number of words correctly recalled (mcorrect
) as a function of task timing
and age
categories, their interaction, while accounting for lab and individual-specific variability.
- Using Oehlert (2000) approach
- Identify whether factors are crossed or nested.
- Determine whether factors should be fixed or random.
- Figure out which interactions can exist and whether they can be fitted.
- Are there difference between recall task (i.e.,
timing
)? - Report estimated marginal means across age groups (separately for each
timing
if there is an interaction), with standard errors. - Report the lab-specific variability and comment on regional differences based on the predictions of the random effects.1
- Compute the correlation of measurements for different individuals in a given lab, following the example 12.4 from the course notes
References
Footnotes
In SPSS, add
SOLUTION
to yourRANDOM
specification for thelab
effect. In R, assuming you fit the model with thelmerTest
orlme4
package, we can extract the predictions of the lab specific mean differences from a fitted model vialme4::ranef
function.↩︎