brant
命令简介
brant是一款进行Stata有序logistic回归平行性检验的Stata社区命令。
代码
另存为brant.ado[1]，并另存为ado\plus\b目录。
*! version 4.0.0 2017-08-23 | long freese | stata 15 parameter naming fix
// brant test of parallel reg assumption
capture program drop brant
program define brant, rclass
version 11
syntax [, DETAILs]
tempvar touse dummy
tempname bout d pvals ivchi ivout step1 step2 ologit
tempname XpWmmX iXpWmmX XpWmlX XpWllX iXpWllX DB DBp iDvBDp
tempname varb bstar id negid zero feed DB DBp iDvBDp step1 step2
tempname step1 wald waldout pout dfout dtemp vbtemp bstemp outmat outvec
if "`e(cmd)'"!="ologit" {
di as error "brant can only be used after ologit"
exit 9991
}
if "`e(wtype)'"!="" {
di as error "brant does not work with weights"
exit 9992
}
local depvar "`e(depvar)'"
gen `touse' = e(sample)
* 4.0.0 2017-08-23 stata 15 parameter naming fix
if (_caller()>=15) version 15: _rm_modelinfo2
else _rm_modelinfo2
local rhsnms "`r(rhsnms)'"
local rhsN "`r(rhsn)'"
local catsN `r(lhscatn)'
local catsNm1 = `catsN' - 1
local catsNm2 = `catsN' - 2
local catvals "`r(lhscatvals)'"
* re-parse ologit model to get syntax for binary logits
_getologitIV // program at bottom of file
local logitvarlist "`s(varlist)'"
_estimates hold `ologit'
// estimate series of binary logits
local i = 1
local estlist ""
while `i' < `catsN' {
local splitat : word `i' of `catvals'
capture drop `dummy'
qui gen `dummy' = 0 if `depvar' <= `splitat' & `touse'==1
qui replace `dummy' = 1 if `depvar' > `splitat' & `touse'==1
qui logit `dummy' `logitvarlist' if `touse' == 1
_rm_modelinfo2
local binrhsN = `r(rhsn)'
if `rhsN' != `binrhsN' {
di as error ///
"not all independent variables can be retained in binary logits"
di as error "brant test cannot be computed"
exit 9993
}
tempvar prob`i'
qui predict `prob`i''
tempname b`i' V`i' bc`i'
mat `b`i'' = e(b)
_rm_matrix_noomitted `b`i'' row
_rm_matrix_noomitted `b`i'' col
mat `b`i'' = `b`i''[1, 1..`rhsN']
mat `V`i'' = e(V)
_rm_matrix_noomitted `V`i'' row
_rm_matrix_noomitted `V`i'' col
mat `V`i'' = `V`i''[1..`rhsN', 1..`rhsN']
mat `bc`i'' = e(b) /* with constant--for detail output only */
mat `bc`i'' = `bc`i'''
_rm_matrix_noomitted `bc`i'' row
_rm_matrix_noomitted `bc`i'' col
local outname "y>`splitat'"
local outname "y_gt_`splitat'"
estimates store `outname'
local estlist "`estlist'`outname' "
mat `bout' = nullmat(`bout'), `bc`i''
local ++i
}
mat rownames `bout' = :
// make variables for W(ml) matrices
local i = 1
while `i' < `catsN' {
local i2 = `i'
while `i2' <= `catsNm1' {
tempvar w`i'_`i2'
qui gen double `w`i'_`i2'' = `prob`i2'' - (`prob`i''*`prob`i2'')
local ++i2
}
local ++i
}
// calculate variance Bm, Bl
local i = 1
while `i' < `catsN' {
local i2 = `i'
while `i2' < `catsN' {
qui {
* inverse(X'W(mm)X)
mat accum `XpWmmX' = `logitvarlist' [iw=`w`i'_`i''] if `touse'==1
_rm_matrix_noomitted `XpWmmX' row
_rm_matrix_noomitted `XpWmmX' col
mat `iXpWmmX' = inv(`XpWmmX')
* X'W(ml)X
mat accum `XpWmlX' = `logitvarlist' [iw=`w`i'_`i2''] if `touse'==1
_rm_matrix_noomitted `XpWmlX' row
_rm_matrix_noomitted `XpWmlX' col
* inverse(X'W(ll)X)
mat accum `XpWllX' = `logitvarlist' [iw=`w`i2'_`i2''] if `touse'==1
_rm_matrix_noomitted `XpWllX' row
_rm_matrix_noomitted `XpWllX' col
mat `iXpWllX' = inv(`XpWllX')
* product of three matrices
mat `step1' = `iXpWmmX'*`XpWmlX'
tempname vb`i'_`i2'
mat `vb`i'_`i2'' = `step1'*`iXpWllX'
}
mat `vb`i'_`i2''= `vb`i'_`i2''[1..`rhsN',1..`rhsN']
local ++i2
}
local ++i
}
// define var(B) matrix
local i = 1
while `i' < `catsN' {
tempname row`i'
local i2 = 1
while `i2' <= `catsNm1' {
if (`i'==`i2') mat `row`i'' = nullmat(`row`i''), `V`i''
if (`i'<`i2') mat `row`i'' = nullmat(`row`i''), `vb`i'_`i2''
if (`i'>`i2') mat `row`i'' = nullmat(`row`i''), `vb`i2'_`i'''
local ++i2
}
local ++i
}
// combine matrices
local i = 1
while `i' < `catsN' {
mat `varb' = nullmat(`varb') \ `row`i''
local ++i
}
// make beta vector
local i = 1
while `i' < `catsN' {
mat `bstar' = nullmat(`bstar'), `b`i''
local ++i
}
mat `bstar' = `bstar''
// create design matrix for wald test; make I, -I, and 0 matrices
local dim = `rhsN'
mat `id' = I(`dim')
mat rownames `id' = `rhsnms'
mat colnames `id' = `rhsnms'
mat `negid' = -1*`id'
mat rownames `negid' = `rhsnms'
mat colnames `negid' = `rhsnms'
mat `zero' = J(`dim', `dim', 0)
mat rownames `zero' = `rhsnms'
mat colnames `zero' = `rhsnms'
* dummy mat
local i = 1
while `i' <= `catsNm2' {
tempname drow`i'
local i2 = 1
while `i2' <= `catsNm1' {
if (`i2'==1) mat `feed' = `id'
else if (`i2'-`i'==1) mat `feed' = `negid'
else mat `feed' = `zero'
mat `drow`i'' = nullmat(`drow`i'') , `feed'
local ++i2
}
local ++i
}
local i = 1
while `i' <= `catsNm2' {
mat `d' = nullmat(`d') \ `drow`i''
local ++i
}
// terms of wald test
mat `DB' = `d'*`bstar'
mat `DBp' = `DB''
mat `step1' = `d'*`varb'
mat `step2' = `step1'*`d''
mat `iDvBDp' = inv(`step2')
// calculate wald stat
mat `step1' = `DBp'*`iDvBDp'
mat `wald' = `step1'*`DB'
sca `waldout' = `wald'[1,1]
sca `dfout' = `rhsN'*`catsNm2'
sca `pout' = chiprob(`dfout', `waldout')
local i = 1
while `i' <= `rhsN' {
tempname d`i' vb`i' bstar`i'
local i2 = 1
while `i2' < `catsN' { // -1
local row = (`rhsN'*(`i2'-1)) + `i'
tempname drow vbrow
local i3 = 1
while `i3' <= `catsNm1' {
local column = (`rhsN'*(`i3'-1)) + `i'
if `i2' < `catsNm1' {
mat `dtemp' = `d'[`row',`column']
mat `drow' = nullmat(`drow') , `dtemp'
}
mat `vbtemp' = `varb'[`row',`column']
mat `vbrow' = nullmat(`vbrow') , `vbtemp'
local ++i3
}
if (`i2'<`catsNm1') mat `d`i'' = nullmat(`d`i'') \ `drow'
mat `vb`i'' = nullmat(`vb`i'') \ `vbrow'
mat `bstemp' = `bstar'[`row', 1]
mat `bstar`i'' = nullmat(`bstar`i'') \ `bstemp'
local i2 = `i2' + 1
}
local ++i
}
// wald test for each independent variable
tempname waldiv
local i = 1
while `i' <= `rhsN' {
mat `DB' = `d`i'' * `bstar`i''
mat `DBp' = `DB''
mat `step1' = `d`i''*`vb`i''
mat `step2' = `step1' * `d`i'''
mat `iDvBDp' = inv(`step2')
mat `step1' = `DBp' * `iDvBDp'
mat `waldiv' = nullmat(`waldiv') \ (`step1' * `DB')
local ++i
}
if "`details'"=="details" {
estimates table `estlist', b(%8.3f) t(%8.2f) ///
title(Estimated coefficients from binary logits)
}
estimates drop `estlist'
local rspec "&--"
mat `outmat' = (`waldout', `pout', `dfout')
mat rowna `outmat' = All
mat colna `outmat' = chi2 "p>chi2" df
local twid = 12
* p for individual wald tests
mat `pvals' = J(`rhsN', 1, 0)
local i = 1
local df = `catsNm2'
while `i' <= `rhsN' {
sca `ivchi' = `waldiv'[`i',1]
if (`ivchi'>=0) mat `pvals'[`i',1] = chiprob(`df',`ivchi')
if (`ivchi'<0) mat `pvals'[`i',1] = -999
local vnm : word `i' of `rhsnms'
local vlen = strlen("`vnm'") + 1
if (`vlen'>`twid') local twid = `vlen'
local rspec "`rspec'&"
mat `outvec' = `ivchi', `pvals'[`i',1], `df'
mat rowna `outvec' = `vnm'
mat `outmat' = nullmat(`outmat') \ `outvec'
local ++i
}
local cspec "o1& %`twid's | %10.2f & %9.3f & %6.0f &"
matlist `outmat', title(Brant test of parallel regression assumption) ///
cspec(`cspec') rspec(`rspec')
di _new ///
"A significant test statistic provides evidence that the parallel"
di "regression assumption has been violated."
mat `ivout' = `waldiv', `pvals'
mat rownames `ivout' = `rhsnms'
mat colnames `ivout' = chi2 p>chi2
return scalar chi2 = `waldout'
return scalar p = `pout'
return scalar df = `dfout'
return matrix ivtests `ivout'
_estimates unhold `ologit'
end
program define _getologitIV, sclass
local 0 "`e(cmdline)'"
local 0 : subinstr local 0 "ologit" ""
version 11
syntax varlist(ts fv) [if] [in] ///
[fw pw iw aw] [, ///
FROM(string) noLOg ///
OFFset(varname numeric) /// -ml model- options
TECHnique(passthru) VCE(passthru) LTOLerance(passthru) ///
TOLerance(passthru) noWARNing ///
Robust CLuster(passthru) /// old options
CRITtype(passthru) SCORE(passthru) ///
DOOPT /// NOT DOCUMENTED
notable /// -Replay- options
noHeader NOCOEF OR ///
* /// -mlopts- options
]
local dv : word 1 of `varlist'
local varlist: subinstr local varlist "`dv'" ""
sreturn local varlist "`varlist'"
end
exit
* version 3.3.1 2017-08-02 | long freese | stata 15 renaming fix
* version 3.3.0 2014-02-14 | long freese | spost13 release