Skip to content

Cook's distance for least absolute deviation (lad) regressions

3 messages · Kelly Thompson, Jim Lemon, Bert Gunter

Kelly Thompson
# 
I'm wanting to calculate Cook's distance for least absolute deviation
(lad) regressions.

Which R packages and functions offer this?

Thanks!
Jim Lemon
# 
Hi Kelly,
Perhaps the best place to look is the "car" package. There is a
somewhat confusing reference in the "cookd" function help page to the
"cooks.distance" function in the "base" package that doesn't seem to
be there.  Whether this is the case or not, I think you can still use
the "cookd" alias.

Jim
On Mon, Mar 21, 2022 at 11:57 AM Kelly Thompson <kt1572757 at gmail.com> wrote:
Bert Gunter
# 
1. See ?influence measures for info on cook's distance in R.

2. See references therein for details; or here for the definition FOR
LEAST SQUARES REGRESSION, but *not lad regression*:
https://www.statisticshowto.com/cooks-distance/
Strictly speaking, Cook's D applies only to L2 regression.

3. I know of no equivalent to Cook's distance for LAD's regression.
Indeed, isn't the whole point of LAD that it is relatively insensitive
to unusual values (which is one of the components of a point's
influence)? However, just because I don't know of it doesn't mean
anything. You might have a look at the 'quantreg' package or search on
"L1 regression in R" for other packages that implement LAD regression
to see what they have. A web search on "influence measures for L1
regression" might also turn up something relevant.


Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
On Sun, Mar 20, 2022 at 6:41 PM Jim Lemon <drjimlemon at gmail.com> wrote: