----- Original Message -----
From: "Clive Nicholas" <[email protected]>
To: <[email protected]>
Sent: Thursday, November 13, 2003 5:48 PM
Subject: st: Turning -mfx- into -MFI-...
> All,
>
> Not for the first time (and certainly not for the last), I'm confused. As
> Max Bygraves would say, "I wanna tell you a stoooory...."
>
> After running -mfx compute- to an -xtprobit- model which several years to
> run (well, that's what it felt like), I left the office to go home, still
> not having finished 2 1/2 hours in.
>
> Fifteen (yes, folks: 15!) hours later, I returned to the office not only
> to find that the locks to all of our office doors had been changed
> (requiring me to call out the university police to let me in), but I
> finally entered my office to discover that Stata *still* hadn't run this
> blessed command. Natch, I was a bit miffed, especially since -mem: 15m,
> matsize: 1000- and nothing else on my machine was left running.
>
> After hitting -Break- (which itself took a fair bit of time to do), I was
> presented with this:
>
> . mfx compute
> fail in calculating second derivative, no standard deviation reported
>
> Marginal effects after xtprobit
> y = Linear prediction (predict)
> = -1.1790009
> ----------------------------------------------------------------------------
> variable | dy/dx X
> ---------------------------------+------------------------------------------
> con00le*| 1.26761 .102012
> lab00le*| -.5775829 .092991
> ldm00le*| .11818 .067314
> tout00le | -.0219548 -.609299
> black*| -8.594328 .008328
> asian*| -.5212507 .020819
> ethother*| -9.29377 .004164
> rghclg97 | -.0250932 2.65996
> tea97 | .0052828 17.9778
> catholic*| -.1466399 .093685
> othxtian*| -.0920777 .120749
> nonxtian*| .2862525 .026371
> norelig*| -.343042 .322693
> marstt97 | -.0560657 2.20402
> tusa297 | .2117155 2.55795
> rsex01 | .0754785 1.50729
> rage01 | .0021773 52.9542
> persno97 | -.07423 1.88827
> larent01*| -.6653879 .11728
> harent01*| -.2289604 .038862
> pvrent01*| -.0678031 .043026
> rtfree01*| .6999252 .008328
> ldclsc01 | -.3029821 2.58501
> conldf01 | .134197 1.94171
> dfwnge01 | -.1203848 2.24566
> eec01 | -.0830513 2.52186
> eurorf01 | .0272745 2.01041
> eurflk01 | -.1604195 1.37266
> rtxspd01 | .0535553 4.07009
> contxs01 | -.06595 6.73907
> labtxs01 | .0435904 4.51076
> ldtxs01 | -.0122722 5.34212
> recind01 | .081757 7.2873
> coneci01 | .0280831 8.29632
> labeci01 | -.1083519 4.83067
> ldeci01 | -.003023 7.01041
> rprfw01 | .0077771 5.39695
> conpr01 | -.0163889 8.09715
> labpr01 | .0208762 6.13393
> ldpr01 | .0066025 8.69396
> gecpst01 | .2016125 2.78557
> gecxpc01 | .122634 2.96461
> ghinps01 | .0463794 2.87925
> ghinxp01 | .0759089 2.92228
> taxspd01 | -.026723 2.59264
> gnhssp01 | .1032135 1.44275
> gvedsp01 | .037254 1.57252
> rdist201 | .2601457 2.6787
> stfsn201 | .009266 1.89313
> wabort01 | .0843802 2.02915
> imopen01 | .0005638 2.81402
> im2mny01 | -.0317217 2.40666
> im2ndc01 | -.0060941 2.97918
> imhome01 | .0095131 3.61207
> gvtrst01 | -.0651358 2.82304
> mptrst01 | .1249612 3.39139
> talkfm01 | -.13457 2.23387
> talkfr01 | -.0671215 2.40597
> talknh01 | .0234844 3.36155
> talkwk01 | .029282 3.33796
> leftri01 | .001262 39.5635
> censor01 | .0119574 2.35253
> tolern01 | .017096 2.47675
> pubmee01 | .064866 2.26926
> hhinca01 | .0026374 15.3428
> inlondon*| -.9364664 .022207
> manchstr*| -.045066 .038168
> mersey*| -.5098787 .018737
> syorks*| -.2012514 .037474
> tynewear*| .2499731 .034698
> westmids*| -.2034792 .058293
> wyorks*| -.0890329 .032616
> avonglos*| -.0702495 .040944
> bedsbuck*| .519918 .022207
> berks*| -.0479753 .012491
> cantab*| .4828031 .020819
> cheshire*| -.0466644 .022901
> clevdurm*| -.3647604 .031228
> corndevn*| -.203371 .03331
> cumbria*| .0694427 .018737
> derbys*| -.1454132 .02984
> dorset*| .6855309 .022901
> esussex*| .0500557 .016655
> essex*| .1747218 .022207
> hants*| .3412985 .016655
> hereworc*| .0678038 .015267
> herts*| .0235927 .03678
> humblinc*| .1528012 .020125
> kent*| -.2571271 .026371
> lancs*| -.0637454 .018043
> leics*| .48924 .026371
> norfolk*| .6963252 .023595
> nyorks*| .17375 .040944
> notts*| -.4579414 .013185
> oxon*| .365233 .027065
> salop*| -.692515 .014573
> staffs*| .9105996 .018043
> suffolk*| .360477 .019431
> surrey*| .0331409 .027065
> warks*| .1745381 .012491
> wsussex*| .5010424 .020819
> wilts*| .4226291 .014573
> ----------------------------------------------------------------------------
> (*) dy/dx is for discrete change of dummy variable from 0 to 1
>
> Now forgive me for being a bit thick (or, indeed, very), but how is it
> that several of these marginal effects *on the probabilities* are greater
> than 1? Is the failure of no second-derivative calculation a real problem,
> here? Or shall I trundle on down to my local supermarket and see what
> Elvis has to say about it all?!
>
Take a look at the examples given in [R] -mfx compute- or in the help
ile. -mfx compute- after -xtprobit- will simply give you the estimated
coefficients.
I imagine you want -mfx compute, predict(pu0)- for the predicted probabilities.
Also, if you do not want the standard errors (or want to wait) use the -nose-
option. This will speed things up considerably.
Elvis has left the building.
Scott
*
* 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/
