-
Examine the GDS data
for the two types of outliers. Conduct frequency distributions of
the variables (they should all be 0 or 1) and visually scan the data
for specific response patterns that might be problematic (e.g., all
0 or all 1). Conduct an EFA using ULS extraction and direct oblimin
rotation, extracting five variables. Determine whether there are any
variables that might be outliers.
-
Replicate our analysis
to impute missing data for a subsample of the SDQ data. Use the syntax
presented at the bottom of this section to select the subsample of
N=300 and generate a nonrandom missing sample by changing all responses
of 6 on Eng1 to missing. Then use the code presented in the chapter
to impute the missing data and run the EFA from the covariance matrix,
using ULS extraction and direct oblimin rotation.
-
Impute missing data
for the GDS data set, using a similar process as we used for the subsample
of SDQ data. Conduct an EFA on the original data and the imputed version
of the data using ULS extraction and direct oblimin rotation, this
time extracting three variables. Compare the results from the model
with imputed data to those without data imputed. Is there a difference
in the results? Why or why not?
Syntax for exercise
2 above:
*Replicate the selection of the 300;
proc surveyselect data=sdqdata method=srs n=300 out=sdqdata_ss1
seed=39302;
run;
*Generate nonrandom missing sample;
data nonrandom_miss;
set sdqdata_ss1;
if Eng1 =6 then Eng1=.;
run;