I am using arimax model to calculate incremental sales due to promotions. Arima code is as below

**out <- try(Arima(unlist(final_data$total_sales), order=c(1,1,0),seasonal=c(0,0,0), xreg=reg_inp, include.mean=TRUE,include.drift=TRUE,transform.pars=TRUE,fixed=NULL, init=NULL, method="ML",kappa=1e6))**  

Co-Efficients of the arima produces negative values for some predictors and I use these coefficients to calculate incremental sales


So my question is an is there any way to restrict my coefficients to give me only positive numbers?

  • $\begingroup$ I appreciate this is an ARIMA model rather than multilinear regression, but it strikes me that you are facing many of the same issues as in this question. $\endgroup$
    – Silverfish
    Commented Feb 25, 2016 at 12:05
  • $\begingroup$ yes but that is on SPSS and on linear regression I believe, but I cannot see any answers to that, all the answers are regarding why negative coefficients are coming up instead of applying restrictions on coefficients. $\endgroup$
    – vivek
    Commented Feb 25, 2016 at 12:11
  • 1
    $\begingroup$ The model you are fitting is not ARIMAX but rather regression with ARMA errors; see Rob J. Hyndman's blog post "The ARIMAX model muddle". $\endgroup$ Commented Feb 25, 2016 at 12:14
  • $\begingroup$ Coding questions are off-topic here - this site really focuses on the underlying statistical issues, rather than software-specific ones. As a result, when searching this site for similar examples, it's usually a mistake to only look for posts where the author was using the same software as you: even if someone is using a different package, they may be facing similar statistical issues. $\endgroup$
    – Silverfish
    Commented Feb 25, 2016 at 12:19
  • $\begingroup$ The reason you can't see any "answers" on the other "forcing the sign of the coefficients" thread that explain how to force the signs is explained very clearly in Nick Cox's comment there. (I wonder if he could be persuaded to post his comment as an answer.) $\endgroup$
    – Silverfish
    Commented Feb 25, 2016 at 12:20


Your Answer

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

Browse other questions tagged or ask your own question.