topic Re: proc nlin - distribution of square error in New SAS User

proc nlin - distribution of square error

linlin87 — Thu, 04 Aug 2022 10:46:27 GMT

Hi SAS Community,

I use proc nlin to fit various models to data.

I want to be able to say for example "if the distribution of square errors has these features, then do ..."

Does proc nlin output the distribution of squared errors as a dataset, or do I have to go back and use the model parameters to manually get the SE distribution?

I am trying make this as efficient as possible as have multiple datasets to process, with each dataset containing ~60,000 rows, and each dataset required ~1,700 non-linear fits (/calls of proc nlin).

Any tips would much be appreciated!

LinLin Jiang

Re: proc nlin - distribution of square error

ballardw — Thu, 04 Aug 2022 17:41:04 GMT

Please at least provide the Proc NLIN code for one of these runs. We need to see the options you are using. It is a headache to suggest "use option X on the PDQ statement" if you don't have a PDQ statement in your code. It may mean that you need to add a PDQ to get the desired result. (PDQ is generic for SOME statement, not an actual NLIN statement).

It is possible that what you want might appear in different tables based on your options and I am not going to make a lot of guesses.

Re: proc nlin - distribution of square error

PGStats — Thu, 04 Aug 2022 18:45:18 GMT

How to get the residuals from proc NLIN, and extra considerations for residuals from non-linear models, is explained here.