r(2000) no observations? I have tried 'destring', but then it says 'already numeric; no generate

[Marie Christine Andreasen]

Hello
I am trying to make a regression using mlogit.
But once I enter a specific one of my variables, it says r(2000) no observations. I have tried using the desting command, but then it says that my variables are already numberic? And now I don't know what to do

Best Regards Marie
A helpless student

[Joro Kolev]

Check out -dataex- and post some minimal example of data that generates the problem using -dataex-.

This error might be caused because the problematic variable is a string, or because the problematic variable has too many missing values.

We cannot tell you unless you show us a subset of your data that generates the problem.

[Phil Bromiley]

To expand on Joro's insightful comment, if the variable happened to have missing values for all the cases where the other variables have non-missing values, you could end up with no usable cases. This can happen naturally but could also happen if you don't get the if conditions right on a generate statement.

You might also want to run a regression and see what you get. You could also could run the mlogit with all the variables that work, then generate a variable that indicates whether the observation was used:
gen used=e(sample)

Then look at the values of the additional variable for observations where used==1. It is often good practice to look at the data.