Missing Data/Imputation Discussion > SPSS Multiple Imputation and Interaction terms
I'm a bit confused. In your example (e.g. A*B=AB), are A & B both examples of the "item/trial level" data you are referring to?
If so, I would likely calculate the interaction terms, use the interaction terms to impute missing items and then calculate interaction terms again to add the interaction terms for the newly imputed data (the data that was already there will not change during this re-calculation, since those items values did not change). I hope this is helpful and please let me know if I misunderstood your question.
July 10, 2013 |
Jeremy Taylor
Jeremy Taylor
Hi! I am using multiple imputation in SPSS for the first time. I am confused about how to handle interaction terms. I believe that I am supposed to include interaction terms in the imputation model (so, I should compute interaction terms like A*B = AB before running MI), but I think I am also supposed to impute missing data at the item/trial level--this trial level data is needed to calculate more accurate interaction terms. So, in summary, I need the imputed trial-level data to calculate the interaction terms I'll use in my main analyses, but I need the interaction terms to impute the trial-level data. Huh? Please let me know if you have suggestions for how to proceed!