0
$\begingroup$

I have taken a survey. Dependent variable is on Likert scale (1-5). Independent variables are also on Likert scale 1-5. What kind of regression should i better use?

I thought about multinomial (or ordinal) logistic regression. can I use it if independent variables are ordinal?

If I use multinomial, should i put ordinal variables into factors or covariates?

Also I tried to use decision trees with 3 independent variables. But in results the depth is only 1 for 1 variable. It's my first time working in SPSS.

$\endgroup$
2
  • $\begingroup$ I would start with ordinal logistic if I were you (I don't see why would you use multinomial if you have an ordinal dependent). With a numerical Likert outcome regular linear regression may also sometimes be fine. $\endgroup$
    – Sointu
    Commented Apr 8 at 11:03
  • $\begingroup$ Linear regression is not recommended for discrete ordinal data that is not interval-scaled. Use a method tailed to ordinal data. $\endgroup$ Commented Apr 8 at 16:18

1 Answer 1

1
$\begingroup$

I would certainly start with ordinal logistic. For the independent variables, ordinal variables are tricky. One tool that I have used a bit, with some success, is optimal scaling. I don't know if it is available in SPSS.

For the tree, the fact that the depth was only 1 may indicate that your independent variables are not strongly related to the dependent variable. That ought to also be apparent in the logistic. However, trees can be fickle with small (or medium) sample sizes. I think Frank Harrell has posted a bunch about that here, but I may have the wrong person.

$\endgroup$
1
  • $\begingroup$ Thank you! I will search for optimal scaling in SPSS $\endgroup$
    – TylerJ
    Commented Apr 8 at 11:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.