lmrob ?

[Christophe Meyer]

Hello,

I use the lmrob() function from the robustbase library in R, to do robust linear regression (i.e. low weights for outliers).

I searched the julia mailing lists and packages list with keywords "lmrob" and "robust", but I didnt find what I'm looking for.

Is there a robust regression available in Julia ?

Meanwhile, I'm looking into RCall.

Regards,

C. M.

[Tom Short]

I'm not aware of anything. The Python package Statsmodels has some well-documented, clean code that looks like it would be easy to port:

https://github.com/statsmodels/statsmodels/tree/master/statsmodels/robust

Their license is compatible with the MIT license, too (R's is not).

--
You received this message because you are subscribed to the Google Groups "julia-stats" group.
To unsubscribe from this group and stop receiving emails from it, send an email to julia-stats...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Christophe Meyer]

Thanks for the pointer. I'm willing to contribute lmrob(), do you think this would be best done as a PR into GLM, another package or a new one ?

[John Myles White]

FWIW, I think the best first step is always to develop your own package. It's pretty easy to merge one package into another. It's much harder to get someone to give you commit access to their package until they've seen some of your work.

 -- John

[Douglas Bates]

On Friday, February 6, 2015 at 3:27:04 PM UTC-6, Christophe Meyer wrote:

Thanks for the pointer. I'm willing to contribute lmrob(), do you think this would be best done as a PR into GLM, another package or a new one ?

In the documentation for the RCall package one of the examples is fitting an lmrob model in R and accessing the fitted model in Julia.