On Wed, 26 Dec 2012, artur tarassow wrote: > despite the fact that I activated "set warning off", > I receive the message "No observations would be left!" after > selecting an obviously non-existing panel data sample. > Actually I would expect no message at all in this case. Is > this intended, or a 'bug'? It is a bit annoying as it > appears in the txt-output using the 'outfile' command.
>From "help set": <quote> warnings: off or on (the default). Suppress or resume the printing of warning messages issued when arithmetical operations produce non-finite values. </quote> So this setting doesn't claim to do what you supposed. Besides, trying to set a null sample is an error, and by design you can't turn _error_ messages off. I'm assuming you're using the --restrict option with the "smpl" command. If you're doing that in an automated fashion in a script you should first check that the command will work, as in <hansl> nulldata 20 set echo off set messages off sample_min = 5 # your choice series condition = uniform() > 0.9 sample_size = sum(condition != 0) if sample_size >= sample_min smpl condition --restrict # and do stuff endif </hansl> Allin Cottrell
