Hi everyone,
I have some questions regarding how far my conclusions can get with statistics procedures (like factor analysis) on tables that had numbers imputed where it were missing:
Lets supose there is a data set with 10000 lines (N of participants) and 5 coluns (variables A, B, C, D, E). Every 200 people lack two of the variables at random (the first 200 answered A, B and C only; the next 200 answered C, D and E only; the next 200 answered B, D and E only; and so on).
What is the difference between filling these gaps using 'expectation maximization' and 'multiple imputation'?
A factor analysis (or correlation matrix) based on the filled table would be reliable? What are its limitations? And, if possible, how can it be measured?
Thanks a lot for every help i can get!