Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have started looking into robust methods of data analysis. In particular, robust regression methods for ANCOVA type models using the "lmRob" function from the "robust" library in R. I'm having a hard time finding what assumptions I need to satisfy in order to use this. I've read that they can be useful in situations where there are outliers, heterogeneity of variance and for non-normal distributions, i.e., (more recently However, I'm unsure of how much departure robust measures will allow for and what other traps might be awaiting me.
References and/or examples will be much appreciated.

Data description:
sample size ~ 1000, differs substantially between groups
1st factor has 3 levels; 2nd factor has 3 levels
3 covariates, 1 of which is related to both factors, 1 related to only one factor

The model I'm especially interested in examining is something like this:

lmr1 <- lmRob(dv ~ cov1 + cov2 + cov3 + fac1 * fac2, data)
share|improve this question
That the residuals from the good part of the data are approximately symmetric, other than that you're safe! PS: you might wanna link to a more recent article, there as been gigantic leaps in computational tractability in this filed in the last 10 years (specially in higher dimensions).... – user603 Oct 28 '11 at 7:06
Thanks. Will look for a more recent paper to update the question. – Matt Albrecht Oct 31 '11 at 4:32

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.