Least Trimmed Squares

[Kwadwo Boateng]

Hi Does anyone have an idea how to run a Least Trimmed Squares regression on stata please.

Many thanks in advance.

[William Lisowski]

The output of search least trimmed squares provides links to two user-written packages that claim to provide this capability.

[Kwadwo Boateng]

Hi William, I looked at what you requested but not much of a help. Any more you could suggest pls?

[William Lisowski]

I know nothing about least trimmed squares regression.

You will perhaps improve improve your chances of obtaining further advice by providing one or more literature references that describe the technique you want to implement. Apparently your needs are not met by either the scls or the robreg user-written commands that the search command I recommended tells us are available from SSC.

[Steve Samuels]

One of the commands yielded by the search that William suggested is robreg (at SSC). Why is the least trimmed squares option (lts) "not much help" to you?

[Kwadwo Boateng]

Hi Steve, I was finally able to find the robreg and check it out. However as per my results below after running my regression, it provides me with no standard errors,t stats nor f stats.

what could be the reason pls?

LTS regression Number of obs = 1,880
Breakdown point = .5
Subsamples = 500
Scale estimate = .46910489

------------------------------------------------------------------------------
Fair | Coef.
-------------+----------------------------------------------------------------
age | .0061593
exp | -.015079
gender | -.1914166
_cons | -.0382071

[Steve Samuels]

what could be the reason pls?

Don't know. The help even says "Computation of standard errors is not supported for LMS, LQS, and LTS." In Stata, your best alternatives look like mmregress or the mm option of robreg. R has a couple of implementations of LTS, but I've not tried them.
\

[Kwadwo Boateng]

Will try it out. Thank you.