Hi,

I am currently puzzled by having a different observation and cluster count in the regression model (4,798 obs. in 1,236 clusters) than when I count the predicted values after exporting in excel (8,016 obs. in 1,652 clusters). After searching on the forum, my first idea was to check if the issue has something to do with missing values in any of the variables used in the regression model. However, even after counting each row that has at least 1 missing observation, I can't explain the difference (as only approx. 100 rows had missing values of at least one variable used in the model).

I used the following commands:

The regression I used follows the following pattern:
Code:
. reg Y A##c.B##c.C D i.E, cluster(ID)
And directly after the regression, I used the prediction command
Code:
 predict FullModelPredictions
Then, I exported the all observations and variables into Excel to more quickly identify patterns with the results explained above.

My intuition is that I should trust the STATA Output. However, I can't explain the difference. Does anyone has an idea what I am missing here?

Best,
Martin