MATLAB Answers

Why does stepwiselm function output NaN with model interactions?

4 views (last 30 days)
Jhon Gray
Jhon Gray on 23 May 2020
Answered: Jeff Miller on 24 May 2020
I was doing stepwise regression and implementing stepwiselm in commant line mdl = stepwiselm(tbl,'interactions'). My dimension of the table was quite big.It's 16 by 88. So there's 87 variable. I put variables in 1 to 87th coulmn and the result is in 88th column. But All the output is comiing NaN. I am not understading why is this happening. The output is comming out one by one but no value.

  0 Comments

Sign in to comment.

Answers (1)

Jeff Miller
Jeff Miller on 24 May 2020
Multivariate techniques like stepwise regression require more cases (rows) than variables--in practice, you need at least 5-10 times as many. Sounds like you have 16 cases and 88 variables, so you unfortunately can't come close to using stepwise. As far as I know, pretty much your only realistic option is look at the 87 predictors one at a time with a univariate regression or correlation. Just be aware that this approach may well tell you--by chance--that some predictors are significant even when they really are not.

  0 Comments

Sign in to comment.