[R] R not giving significance tests for coefficients/estimates?

Thomas Lumley tlumley at uw.edu
Tue Jan 24 20:44:57 CET 2012


On Tue, Jan 24, 2012 at 9:36 AM, Bert Gunter <gunter.berton at gene.com> wrote:
> 1. It's an interaction in a formula not a multiplication.
>
>
> 2. Read the Help file. It says to use ltsreg not lmsreg.
>
> 3. But no P values will be given anyway. The reason is that exact
> distributions and therefore reliable P values or CI's for small
> samples do not exist for these robust procedures (they would
> undoubtedly depend on the algorithm's control parameters, anyway).
> Only asymptotic results (I think).  I believe the bootstrap is often
> used, but that of course may be problematic due to computational
> considerations.


The bootstrap should be perfectly feasible with modern computers.  It
may be worth pointing out that the bootstrap does not work for LMS
regression, even asymptotically, which is another reason to use LTS
instead.

   -thomas

-- 
Thomas Lumley
Professor of Biostatistics
University of Auckland



More information about the R-help mailing list