1

I have been using rlm() in the MASS library in R with the redescending weights (using MM or Tukey's biweight function). I wanted to find the importance of each predictor in the fitted model. Can someone please point me to a resource on how to go about doing this?

I wanted to clarify that I am looking for finding the importance of each predictor in robust linear regression. I am aware of how these work for linear models but not for robust linear regression. The link pointed to does not specifically say if the relaimpo package is valid for robust linear regression.

I am not sure about using bootstrap methods because of the concern that the bootstrap method (which assumes that all observations are i.i.d.) may not hold in the presence of contamination. Therefore, I was preferably looking for a method which works without the bootstrap.

I also wanted to clarify that I am looking for some deeper explanation/understanding on how to go about doing it. Pointing to a R package is fine (even though this forum is not about R) but I am really wanting to understand how this is done.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
user122870
  • 11
  • 1
  • 3
    Welcome to Cross Validated. If you are looking for references for the maths/methodology, please add the `references` tag to your post. If your question is asking for the reference for doing it in R, note that questions that are solely about programming are [off-topic](http://stats.stackexchange.com/help/on-topic) for this site and may be closed – Marquis de Carabas Jul 11 '16 at 06:41
  • At the moment the way the question is phrased does make it hard to know whether this is a reference request, and you are seeking an article or textbook that you can refer to for deeper explanation, or whether you are hoping for full answers here. We generally distinguish between these two kinds of questions. I think you'll get answers that are "better" for what you're looking for, if you could take a few minutes to edit your question and convey more clearly what you're hoping to see. – Silverfish Jul 11 '16 at 20:44
  • 3
    Please register &/or merge your accounts (you can find information on how to do this in the **My Account** section of our [help]), then you will be able to edit & comment on your own question. – Silverfish Jul 11 '16 at 20:55
  • This what you are looking for I suppose - http://www.statmethods.net/stats/regression.html – wololo Jul 11 '16 at 06:40
  • Please register &/or merge your accounts (you can find information on how to do this in the **My Account** section of our [help]), then you will be able to edit & comment on your own question. – gung - Reinstate Monica Jul 12 '16 at 00:41

0 Answers0