Is there a situation where Stata creates new observations when you merge? I ran a model on a master dataset and was able to replicate a published model. Then when I merged data into the master dataset and ran the same model, the N increases by 13 observations. I tested this out by creating an indicator variable for all observations that made it into the sample when I ran the initial model. And when I merge the datasets, there are 13 new observations for this indicator variable. Any thoughts?