I have a datasample with some low-probability dependent binary variable. In order to get better estimates, I want to use "upsampling" and penalize stronger for wrong estimates of the TRUE-class.
In the documentation of the melogit command (https://www.stata.com/manuals/memelogit.pdf ) , it says that it accepts weights.

However, when I use this:

Code:
 melogit y x1 x2 x3 [fw = weights] ||  ID:
where weights is defined in the sample (1 for y=0 and 25 for y=1), I get this message:

Code:
weights not allowed
r(101);

Am I doing something wrong?

Thanks a lot!