Proper way of using weights


I’ve been using PERWT in STATA with the command [fw= perwt] in some datasets, but recently with two specific datasets (Bolivia 2012 and Paraguay 2012) it shows an error because the variable perwt is noninteger: may not use noninteger frequency weights.

What sould I do in this case? It also doesn’t allow me to use pweight when I try to apply the command.


In this case, we recommend rounding the weights to the nearest integer.

Thank you for the answer. In this case i’m just concerned because the perwt’s range is 9.69-10.32, so every value will be rounded to 10. Is there any other solution in this case?

Many thanks in advance.

In Stata, most estimation commands accept sampling weights (pweight). So while you can’t run -tabulate- without rounding the weights, you can run -total- with the “over()” option to get the same numbers, and this will accept pweights.