e(ll) not reported by regress error

[Sanne Bruijniks]

Dear all,

I'm conducting the command gvselect, but it gives me the following error:e(ll) not reported by regress
I do not understand why, I ran the regression again using eretrun list and it gives me the e(ll)

Does anybody have a clue what goes wrong here?

Thanks for thinking along in advance,

Best regards,

Sanne

syntax:

regress BDIm6 BDIbas
predict residchangescore, residuals

ssc install gvselect
gvselect <term> SelfreporteduseofAD NumberofC s1bg1 s1bg2 s1bg3 s1bg4 s1bg5 s1bg6 s1bg7 s1bg9 s1bg10 s1bg11 s1bg13 s1bg14 s1bg15 s1bg17 s1bg18 s1bg19 s1bg20 s1bg25 s1bg28 s1bg32 Happinessbas Moodssbas s1g22 RAND36sociaaltrans RAND36fysiektrans RAND36rolbeperkingfysiektrans RAND36rolbeperkingemotrans RAND36mentalegeztrans RAND36vitatrans RAND36gezvertrans RAND36pijntrans s1eq6 ACMTQaut ACMTQcon CCTStotal totalCCLbaseline totalCCLdepbaseline totalCCLanxbaseline RDQdepsymp RDQomgaanmetstress RDQposgez RDQfunctioneren RDQtevredenheidleven RDQalgemeenwelbevinden IPTskills IPTpreference CGTpreference s1et1 s1et2 s1et3 s1et4 s1et5 s1et6 s1et7 s1et8 s1et9 baselineBADSactivatie baselineBADSvermijdingruminatie baselineBADSwerkschool baselineBADStotal baselineCTQEM baselineCTQFM baselineCTQEV baselineCTQLV baselineCTQSM BDIOZA RAND36baseline BDIbaseline, nmodel(1): regress residchangescore <term>

[Joseph Coveney]

I'm not sure what gvselect is or does, but from the looks of it I'm guessing that it's some kind of throw-everything-into-the-hopper-and-get-me-significance command, and with all of those candidate predictors and your dataset size you have rank 0. Maybe something like this is happening:

.ÿ
.ÿversionÿ15.1

.ÿ
.ÿclearÿ*

.ÿ
.ÿsysuseÿauto
(1978ÿAutomobileÿData)

.ÿquietlyÿregressÿgearÿdisplacement

.ÿdisplayÿinÿsmclÿasÿtextÿe(ll)
-3.4533536

.ÿ
.ÿquietlyÿdropÿinÿ3/l

.ÿregressÿgearÿdisplacement

ÿÿÿÿÿÿSourceÿ|ÿÿÿÿÿÿÿSSÿÿÿÿÿÿÿÿÿÿÿdfÿÿÿÿÿÿÿMSÿÿÿÿÿÿNumberÿofÿobsÿÿÿ=ÿÿÿÿÿÿÿÿÿ2
-------------+----------------------------------ÿÿÿF(1,ÿ0)ÿÿÿÿÿÿÿÿÿ=ÿÿÿÿÿÿÿÿÿ.
ÿÿÿÿÿÿÿModelÿ|ÿÿÿ.55124995ÿÿÿÿÿÿÿÿÿ1ÿÿÿ.55124995ÿÿÿProbÿ>ÿFÿÿÿÿÿÿÿÿ=ÿÿÿÿÿÿÿÿÿ.
ÿÿÿÿResidualÿ|ÿÿÿÿÿÿÿÿÿÿÿ0ÿÿÿÿÿÿÿÿÿ0ÿÿÿÿÿÿÿÿÿÿÿ.ÿÿÿR-squaredÿÿÿÿÿÿÿ=ÿÿÿÿ1.0000
-------------+----------------------------------ÿÿÿAdjÿR-squaredÿÿÿ=ÿÿÿÿÿÿÿÿÿ.
ÿÿÿÿÿÿÿTotalÿ|ÿÿÿ.55124995ÿÿÿÿÿÿÿÿÿ1ÿÿÿ.55124995ÿÿÿRootÿMSEÿÿÿÿÿÿÿÿ=ÿÿÿÿÿÿÿÿÿ0

------------------------------------------------------------------------------
ÿÿgear_ratioÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿtÿÿÿÿP>|t|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-------------+----------------------------------------------------------------
displacementÿ|ÿÿ-.0076642ÿÿÿÿÿÿÿÿÿÿ.ÿÿÿÿÿÿÿÿ.ÿÿÿÿÿÿÿ.ÿÿÿÿÿÿÿÿÿÿÿÿ.ÿÿÿÿÿÿÿÿÿÿÿ.
ÿÿÿÿÿÿÿ_consÿ|ÿÿÿ4.507372ÿÿÿÿÿÿÿÿÿÿ.ÿÿÿÿÿÿÿÿ.ÿÿÿÿÿÿÿ.ÿÿÿÿÿÿÿÿÿÿÿÿ.ÿÿÿÿÿÿÿÿÿÿÿ.
------------------------------------------------------------------------------

.ÿdisplayÿinÿsmclÿasÿtextÿe(ll)
.

.ÿ
.ÿexit

endÿofÿdo-file


.

[Sanne Bruijniks]

Dear Joseph,

Thank you for your response. We use gvselect to select predictors for our prediction model (using AIC). I am not sure what you mean with the model you post, what do you think goes wrong?

Best regards, Sanne

[Joseph Coveney]

Try reducing the number of candidate predictors by a few and see whether it helps.