Are there any issues to run a Newey-West time-series regression on a dependent variable that is a probability? What are the biases that I am facing? I can't find anything online that can help me out with this issue.
$\begingroup$
$\endgroup$
2
-
$\begingroup$ I have to be more precise, it is more of a proportion than a probability for my dependent variable where the value can be between 0 and 1 $\endgroup$– CharlesMCommented Nov 4, 2013 at 18:56
-
$\begingroup$ I could use soemthing like this ats.ucla.edu/stat/stata/faq/proportion.htm but how can I correct for autocorrelation? $\endgroup$– CharlesMCommented Nov 4, 2013 at 19:00
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
For dependent variables distributed continuously between 0 and 1, you'll want to use a beta regression. I don't know what statistical software you are using, but this article should be helpful, especially if you are using R: http://cran.r-project.org/web/packages/betareg/vignettes/betareg.pdf
In order to help control for auto-correlation, try including a lagged version of the outcome variable as a predictor in the model. Then you can run a set of diagnostics to see if auto-correlation is still a problem.
-
$\begingroup$ Thanks for the help. I just saw that I can do a beta regression also in SAS: support.sas.com/resources/papers/proceedings12/325-2012.pdf which is my commonly use software. So if I understand corretly, a beta regression is part of the GLM family $\endgroup$– CharlesMCommented Nov 4, 2013 at 21:29