-1
$\begingroup$

My dependent variable is classified in categories as 5-10, 10-15 . Which is the best regression model for this kind of analysis.my dependent variable asks the participant of the survey to mark the percentage of turnover due to innovations and the options are (1)less than 1% (2) 1%-5% (3)5%-15% (4) 15%-30% and(5) above 30% . Now please please help me to find the correct regression model, should I dummy code them and use them in linear regression or I should dummy code them and use for multinomial. my independent variables are all in likert scale. please help

$\endgroup$
4
  • 1
    $\begingroup$ The question is quite unclear to me. Try adding details regarding the following - - What are these dependent variables representing - are they just random names or is there some ordinality to them? - What is the range of this variable? - Is the dataset balanced if they are classes? - Any kind of special cases like large dataset, high cardinality of features etc? Depending on these questions the answers and best model will change. $\endgroup$ – Axelius Mar 11 '19 at 14:11
  • $\begingroup$ If you have only two levels in the dependent variable then why should it follow multinomial instead of binomial? $\endgroup$ – Digio Mar 11 '19 at 16:22
  • $\begingroup$ My dependent variable is classified in categories as 5-10, 10-15 . Which is the best regression model for this kind of analysis.my dependent variable asks the participant of the survey to mark the percentage of turnover due to innovations and the options are (1)less than 1% (2) 1%-5% (3)5%-15% (4) 15%-30% and(5) above 30% . Now please please help me to find the correct regression model, should I dummy code them and use them in linear regression or I should dummy code them and use for multinomial. my independent variables are all in likert scale. please help $\endgroup$ – Richa Mar 11 '19 at 17:35
  • $\begingroup$ One option could be linear regression with maximum likelihood for an interval censored response $\endgroup$ – kjetil b halvorsen Mar 12 '19 at 12:44
0
$\begingroup$

For the dependent variable, I would start with ordinal logistic regression and see if the assumptions are met. There are various sorts of ordinal logistic, but the most common (and the place I'd start) is proportional odds logistic.

Likert scaled IVs are interesting. There's no common method, but I've sometimes used optimal scaling with some good results.

All then above is quite general - it may not be good with your particular data set. You might want to hire a consultant.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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