Monday, September 17, 2012

Just curious...

I found a funky work-around that does the trick for what I'm working on, but since I've gotten good answers from random questions here before on everything from drills to general solutions to a polynomial with non-integer exponents, I thought I'd ask:

If you do non-linear least squares in Stata for some reason it only seems to save all the information from the last parameter for you to return as a scalar. I don't have a lot of experience writing rclass type programs in Stata (I usually just do normal programming in a do file - none of the programming commands), though, so maybe it does store more than that. But when I hit "return list" it looks like it's just saving the last parameter.

Is that true? Why does it do that?

My work around is that I just rearranged the way that I wrote the non-linear equation so that each parameter gets it's turn as the last parameter. I return all the scalars and run through however many hundreds of simulations I need to. That does the trick. But obviously not all non-linear equations are amenable to that somewhat tedious work around, so I'm curious what you would do in normal situations.

2 comments:

  1. Are you posting things to a postfile? You can post to multiple variables in a postfile. Here's a good explanation: http://datatoday.blogspot.com/2011/11/using-postfile-in-stata.html

    ReplyDelete
  2. Perhaps one might want to use non-parametric statistics?

    ReplyDelete

All anonymous comments will be deleted. Consistent pseudonyms are fine.