Robust regression and singular error

Question

I got this error message from a robust regression in r ; "'x' is singular: singular fits are not implemented in 'rlm'"

I could solve it by removing some variables.

My question is , can i progrom r to solve it on it's own ? so i will not have to remove variables to solve it.

'How to program it' is off topic here. FWIW, the "'x' is singular' means you have multicollinearity. — gung - Reinstate Monica, Nov 16 '15 at 12:46
Simple solution: use the output of lm to find the variables it dropped. Unless rlm is reweighting a relatively large number of observations to zero (which would be hard to do), this will do the trick. — whuber, Jul 29 '20 at 15:24

score 2 · Answer 1 · edited Jul 29 '20 at 15:19

2

I have come across the same issue with rlm() when an x (predictor) variable is coded as a factor with empty levels (for example, factor(c("a", "b"), levels=c("a","b", "c" ))). Simply converting the factor to a character vector solved my issue. Alternatively, one can use forcats::fct_drop().

edited Jul 29 '20 at 15:19

CoderGuy123

437
4
14

answered Jun 22 '18 at 18:20

Jing Xue

121

I had this issue as well. Really hard to find the cause, though I knew it had to do with that factor variable. – CoderGuy123 Jul 29 '20 at 14:33

score 0 · Answer 2 · answered Mar 13 '17 at 22:32

0

Check for NA values, multicollinearity and duplicate observations. multicollinearity For duplicate observations use jitter rnorm() as suggested: Cause of singularity in matrix for quantile regression.

answered Mar 13 '17 at 22:32

ran8

157

Robust regression and singular error

2 Answers2