|
|
Hi all,
I am new at multiple imputation. I have started working on it recently for my dissertation. This is what I don't understand.
1. There are certain tasks/activities on SPSS where I will get pooled data information (e.g., correlations) and certain tasks (e.g., scatter plots, histograms) that do not show any pooled data information. Even when I checked my predictors for multicollinearity, there were 5 rows of VIF and Tolerance reported. How do I choose which dataset to talk about when I need to discuss the graphs for the purposes of reporting floor effects, linearity and reporting multicollinearity? What do experts typically do?
2. In relation to my first question, typically, do researchers choose one MI data set (e.g., last MI data set) if the pooled correlations look like there wasn't much for the multiple data sets? If I chose only one data set, wouldn't that mean that I am no longer conducting a real multiple imputation?
3. Should I then use a footnote to explain that for the scatter plots and histograms I used only one dataset because of the limitation of the software? I wonder how experts in the field handle such situations.
4. Can I just import that dat file into Mplus and Mplus will understand that there were e.g., 5 multiple data sets?
I appreciate your time and advice in sharing what you know. Thank you very much!
|
Locked
