I found this thread very helpful, thanks to all. I was on the verge of reporting a bug because nlogit was inexplicably dropping a large number of records without any warning. I summarized my data to check that there were no missing values and noticed that one of my id variables had some rather large values. It then ocurred to me that maybe Stata was internally producing an overflow with some operation on that variable, which would internally produce missing values and result in records being dropped. I replaced the long id variable with a simple sequential id and re-ran the code. This time it did not drop any data so I conclude that the above issue was indeed what was causing the problem.
This behaviour of Stata is not helpful - ideally it would warn the user if a Stata command (not user-written code) internally produces missing values that result in data being dropped. Could it be considered a bug? I'm not sure.
This behaviour of Stata is not helpful - ideally it would warn the user if a Stata command (not user-written code) internally produces missing values that result in data being dropped. Could it be considered a bug? I'm not sure.
Comment