Monday, January 02, 2006

RE: st: Efficiency of calculating maximum likelihood

It appears that the initial values for the constant-only model are hard-coded.

Using the lognormal distribution as an example, you can alter (after making a copy for safekeeping) -lnormal.ado- so that if accepts the initial values passed to it.

If you alter line 142: */ init(_cons=1) /*

to */ `iniopt' /*

it will take the values passed to it from -from()-

Now, there is only one iteration in the constant-only model in the second specification below:

sysuse cancer, clear qui stset studytime, failure(died) qui streg , dist(lnorm) matrix b0 = e(b) streg study age , dist(lnorm) from(b0) trace

. qui stset studytime, failure(died)

. qui streg , dist(lnorm)

. matrix b0 = e(b)

. streg study age , dist(lnorm) from(b0) trace

failure _d: died analysis time _t: studytime

Fitting constant-only model:

Iteration 0: Coefficient vector: _t: ln_sig: _cons _cons r1 2.768933 .1422957

log likelihood = -60.991749 Iteration 1: Coefficient vector: _t: ln_sig: _cons _cons r1 2.768933 .1422957

log likelihood = -60.991749

Fitting full model: . . .

Hope this helps, Scott

> -----Original Message----- > From: owner-statalist@hsphsun2.harvard.edu [mailto:owner- > statalist@hsphsun2.harvard.edu] On Behalf Of Maarten buis > Sent: Monday, January 02, 2006 2:33 PM > To: statalist@hsphsun2.harvard.edu > Subject: RE: st: Efficiency of calculating maximum likelihood > > Hi Steinar, > > My initial guess was you could use the previous estimates as starting > values. It might speed > things up, but as the example below shows, that does not have to be. In > the example the model with > starting values needed more itterations than the model without. > > *------------begin example-------------- > sysuse cancer, clear > stset studytime, failure(died) > xi: streg i.drug age, dist(lnorm) > matrix b0 = e(b) > xi: streg i.drug , dist(lnorm) from(b0, skip) > xi: streg i.drug , dist(lnorm) > *-------------end example----------------- > > HTH, > Maarten > > > ----------------------------------------- > Maarten L. Buis > Department of Social Research Methodology > Vrije Universiteit Amsterdam > Boelelaan 1081 > 1081 HV Amsterdam > The Netherlands > > visiting adress: > Buitenveldertselaan 3 (Metropolitan), room Z214 > > +31 20 5986715 > > http://home.fsw.vu.nl/m.buis/ > ----------------------------------------- > > Steinar Fossedal wrote: > > I'm trying to fit a model using stepwise selection, and so I > > reestimate fairly similar models over and over. Fitting a model > > using -streg- takes about an hour using my dataset. I noticed > > that two thirds of the calculation time is spent calculating > > the constant-only model - which is exactly the same for all the > > estimations. Is there any way of making stata use the results > > of previous calculations to save time in this process? >

* * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/


Tag:


Links to this post:

Create a Link



<< Home

This page is powered by Blogger. Isn't yours?