Hello everyone!
I have encountered a problem in my analysis for my thesis. Despite having around 2000 observation my regressions indicated only around 1000. I'm working with multiply imputed data and should therefore not have a problem with missing values. I checked anyways to be thorough and found a lot of missing values. This puzzled me a lot since I had checked all my variables along the way. No missing values. After a lot of backtracking I discovered that my data set of mi data where fine. Saved in the "wide" style they showed no missing variables for M=1 to M=5.
Hence, the problem seems to occour during the merge of my data sets (one from 2009 and one from 2010 with different variables). Right after the merge, M=1 through to M=5 showed the same amount of missing values as did M=0, despite showing none before the merge. I'm extremly puzzled and have no idea where I am going wrong.
I used the following commands:
use "file2009.dta", clear
sort respid
mi merge 1:1 respid using "file2010.dta", keep(match)
The 2010 file has also been sorted by respid beforehand. Thank you very much to anyone who can help me.
Greetings Eva
0 Response to Mi merge problem
Post a Comment