You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
890 lines
30 KiB
Plaintext
890 lines
30 KiB
Plaintext
*! version 2.4 june2020
|
|
*! Myriam Blanchin - Priscilla Brisson
|
|
************************************************************************************************************
|
|
* ROSALI: RespOnse-Shift ALgorithm at Item-level
|
|
* Response-shift detection based on Rasch models family
|
|
*
|
|
* Version 1 : December 21, 2016 (Myriam Blanchin) /*rspcm122016*/
|
|
* Version 1.1 : October 13, 2017 (Myriam Blanchin) /*option: MODA, automatic recoding of unused response categories*/
|
|
* Version 2 : April, 2018 (Myriam Blanchin - Priscilla Brisson) /*option: GROUP, dichotomous group variable*/
|
|
* Version 2.1 : October, 2018 (Myriam Blanchin - Priscilla Brisson) /* Version 1.1 + Version 2 */
|
|
* Version 2.2 : February, 2019 (Priscilla Brisson) /* option nodif, optimization */
|
|
* Version 2.3 : December, 2019 (Priscilla Brisson) /* option detail, + petites corrections */
|
|
* Version 2.4 : June, 2020 (Myriam Blanchin) /* debug option detail + step C, modifs sorties et help */
|
|
*
|
|
* Myriam Blanchin, SPHERE, Faculty of Pharmaceutical Sciences - University of Nantes - France
|
|
* myriam.blanchin@univ-nantes.fr
|
|
*
|
|
* Priscilla Brisson, SPHERE, Faculty of Pharmaceutical Sciences - University of Nantes - France
|
|
* priscilla.brisson@univ-nantes.fr
|
|
*
|
|
* This program is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
* (at your option) any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program; if not, write to the Free Software
|
|
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
************************************************************************************************************/
|
|
|
|
program define rosali_nobf, rclass
|
|
|
|
timer clear 1
|
|
timer on 1
|
|
|
|
syntax varlist(min=2 numeric) [if] [,GROUP(varlist) NODIF PRO DETail]
|
|
|
|
preserve
|
|
version 15
|
|
tempfile saverspcm
|
|
capture qui save `saverspcm',replace
|
|
local save1=_rc
|
|
|
|
if "`if'"!="" {
|
|
qui keep `if'
|
|
}
|
|
|
|
if "`pro'" != "" {
|
|
di "START"
|
|
}
|
|
|
|
/**************************************************************************/
|
|
set more off
|
|
set matsize 5000
|
|
constraint drop _all
|
|
|
|
local gp "`group'"
|
|
|
|
tokenize `varlist'
|
|
local nbitems:word count `varlist'
|
|
|
|
/* Vérif nb d'items pair */
|
|
local mod=mod(`nbitems',2)
|
|
if `mod'!=0 {
|
|
di as error "You must enter an even number of items : the first half of the items represents the items at time 1 and the second half the items at time 2"
|
|
error 198
|
|
exit
|
|
}
|
|
|
|
local nbitems=`nbitems'/2
|
|
|
|
|
|
if "`group'"=="" & "`nodif'"!="" {
|
|
di as error "nodif can only be used with the group option ({hi:nodif} option). Please correct this option."
|
|
error 198
|
|
exit
|
|
}
|
|
|
|
local nbc: word count `group'
|
|
if `nbc' >= 2 {
|
|
di as error "Only one variable can be used for group option ({hi:group} option). Please correct this option."
|
|
error 198
|
|
exit
|
|
}
|
|
|
|
/* Vérif qu'il y a 2 groupes si l'option groupe est choisie */
|
|
if "`group'"!="" {
|
|
qui tab `group'
|
|
local nbgrp = r(r)
|
|
if `nbgrp' != 2 {
|
|
di as error "Only 2 groups are possible for the group option ({hi:group} option). Please correct this option."
|
|
error 420
|
|
exit
|
|
}
|
|
}
|
|
/* recoder la variable de groupe en 0, 1*/
|
|
|
|
if "`group'"!="" {
|
|
qui tab `gp', matrow(rep)
|
|
qui matrix list rep
|
|
if rep[1,1]+rep[2,1] != 1 & rep[1,1]*rep[2,1] != 0 {
|
|
forvalues i=1/`=rowsof(rep)'{
|
|
qui replace `gp'=`i'-1 if `gp'==rep[`i',1]
|
|
di "WARNING : `gp' `=rep[`i',1]' is now `gp' `=`i'-1' "
|
|
}
|
|
}
|
|
forvalues g = 0/1 {
|
|
qui tab `gp' if `gp' == `g'
|
|
local nbp_gp`g' = r(N)
|
|
}
|
|
}
|
|
|
|
|
|
|
|
/*item rename*/
|
|
/*
|
|
Items au temps 1 : 1 à nbitems ``j''
|
|
Items au temps 2 : nbitems à 2*nbitems ``=`j'+`nbitems'''
|
|
|
|
Si t varie, puis num item : ``=(`t'-1)*`nbitems'+`j'''
|
|
*/
|
|
|
|
|
|
local com_z = 0 // Indicatrice de recodage
|
|
/*verif modalités répondues*/
|
|
if "`gp'" == "" { // Si pas d'option groupe
|
|
forvalues j = 1 / `nbitems' {
|
|
local recoda_`j' = 0
|
|
qui tab ``j'', matrow(rect1_`j') // Récupération des infos moda du temps 1
|
|
local minm`j'_t1 = rect1_`j'[1,1]
|
|
local maxm`j'_t1 = rect1_`j'[r(r),1]
|
|
|
|
qui tab ``=`j'+`nbitems''', matrow(rect2_`j') // Récupération des infos moda du temps 2
|
|
local minm`j'_t2 = rect2_`j'[1,1]
|
|
local maxm`j'_t2 = rect2_`j'[r(r),1]
|
|
|
|
local minm_`j' = min(`minm`j'_t1',`minm`j'_t2') // Info moda pour l'item j
|
|
local maxm_`j' = max(`maxm`j'_t1',`maxm`j'_t2')
|
|
local nbm_`j' = `=`maxm_`j''-`minm_`j'''
|
|
|
|
if `minm_`j'' != 0 & `com_z' == 0 {
|
|
local com_z = 1
|
|
}
|
|
|
|
|
|
//Recodage des réponses en 0, 1, 2, etc...
|
|
forvalues r = 0/`=`maxm_`j''-1' {
|
|
qui replace ``j'' = `r' if ``j'' == `=`r'+`minm_`j'''
|
|
qui replace ``=`j'+`nbitems''' = `r' if ``=`j'+`nbitems''' == `=`r'+`minm_`j'''
|
|
}
|
|
|
|
// Vérif. Que toutes les modas sont utilisées & concordance entre temps
|
|
forvalues m = 0/`nbm_`j'' {
|
|
qui count if ``j'' == `m'
|
|
local nb_rn1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `m'
|
|
local nb_rn2 = r(N)
|
|
local nb_rn = min(`nb_rn1',`nb_rn2')
|
|
|
|
if `nb_rn' == 0 { // Une moda n'est pas utilisée
|
|
local recoda_`j' = 1
|
|
if `m' == 0 | `m' <= `minm`j'_t1' | `m' <= `minm`j'_t2' { // La moda 0 ou les moda min ne sont pas utilisées
|
|
local stop = 1
|
|
forvalues k = 1/`=`nbm_`j''-`m'' {
|
|
qui count if ``j'' == `=`m' + `k''
|
|
local v`k'1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' + `k''
|
|
local v`k'2 = r(N)
|
|
if (`v`k'1' != 0 | `v`k'2' != 0) & `stop' != 0 {
|
|
qui replace ``j''= `=`m'+`k'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''=`=`m'+`k'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `=`m'+`k'' merged "
|
|
local stop = 0
|
|
}
|
|
}
|
|
}
|
|
else if `m' >= `maxm`j'_t1' | `m' >= `maxm`j'_t2' | `m' == `maxm_`j'' { // La (ou les) moda max ne sont pas utilisée(s)
|
|
local stop = 1
|
|
forvalues k = 1/`m' {
|
|
qui count if ``j'' == `=`m' - `k''
|
|
local v`k'1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' - `k''
|
|
local v`k'2 = r(N)
|
|
if (`v`k'1' != 0 | `v`k'2' != 0) & `stop' != 0 {
|
|
qui replace ``j''=`=`m' - `k'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''=`=`m' - `k'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `=`m'-`k'' merged"
|
|
local stop = 0
|
|
}
|
|
}
|
|
}
|
|
else {
|
|
if runiform()>0.5{ // Tirage au sort pour regrouper
|
|
local stop = 1
|
|
forvalues k = 1/`m' {
|
|
qui count if ``j'' == `=`m' - `k''
|
|
local v`k'1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' - `k''
|
|
local v`k'2 = r(N)
|
|
if (`v`k'1' != 0 | `v`k'2' != 0) & `stop' != 0 {
|
|
qui replace ``j''= `=`m'-`k'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems''' =`=`m'-`k'' if ``=`j'+`nbitems''' ==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `=`m'-`k'' merged"
|
|
local stop = 0
|
|
}
|
|
}
|
|
}
|
|
else {
|
|
local stop = 1
|
|
forvalues k = 1/`=`nbm_`j''-`m'' {
|
|
qui count if ``j'' == `=`m' + `k''
|
|
local v`k'1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' + `k''
|
|
local v`k'2 = r(N)
|
|
if (`v`k'1' != 0 | `v`k'2' != 0) & `stop' != 0 {
|
|
qui replace ``j''=`=`m' + `k'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''=`=`m' + `k'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `=`m'+`k'' merged"
|
|
local stop = 0
|
|
}
|
|
else {
|
|
if `stop' != 0 {
|
|
qui replace ``j''= `nbm_`j'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''= `nbm_`j'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `nbm_`j'' merged"
|
|
local stop = 0
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
else { // Cas où l'option groupe est utilisée
|
|
forvalues j = 1 / `nbitems' {
|
|
local recoda_`j' = 0
|
|
qui tab ``j'' if `gp' == 0, matrow(rect1_g0_`j') matcell(nbrt1_g0_`j') // Récupération des infos moda du temps 1pour chaque groupe
|
|
local minm`j'_t1_g0 = rect1_g0_`j'[1,1]
|
|
local maxm`j'_t1_g0 = rect1_g0_`j'[r(r),1]
|
|
|
|
qui tab ``j'' if `gp' == 1, matrow(rect1_g1_`j') matcell(nbrt1_g1_`j')
|
|
local minm`j'_t1_g1 = rect1_g1_`j'[1,1]
|
|
local maxm`j'_t1_g1 = rect1_g1_`j'[r(r),1]
|
|
|
|
qui tab ``=`j'+`nbitems''' if `gp' == 0, matrow(rect2_g0_`j') matcell(nbrt2_g0_`j') // Récupération des infos moda du temps 2 pour chaque groupe
|
|
local minm`j'_t2_g0 = rect2_g0_`j'[1,1]
|
|
local maxm`j'_t2_g0 = rect2_g0_`j'[r(r),1]
|
|
|
|
qui tab ``=`j'+`nbitems''' if `gp' == 1 , matrow(rect2_g1_`j') matcell(nbrt2_g1_`j')
|
|
local minm`j'_t2_g1 = rect2_g0_`j'[1,1]
|
|
local maxm`j'_t2_g1 = rect2_g0_`j'[r(r),1]
|
|
|
|
local minm_`j' = min(`minm`j'_t1_g0',`minm`j'_t2_g0',`minm`j'_t1_g1',`minm`j'_t2_g1') // Info moda pour l'item j
|
|
local maxm_`j' = max(`maxm`j'_t1_g0',`maxm`j'_t2_g0',`maxm`j'_t1_g1',`maxm`j'_t2_g1')
|
|
local nbm_`j' = `=`maxm_`j''-`minm_`j''+1'
|
|
|
|
if `minm_`j'' != 0 & `com_z' == 0 {
|
|
local com_z = 1
|
|
}
|
|
//Recodage des réponses en 0, 1, 2, etc...
|
|
forvalues r = 0/`=`maxm_`j''-1' {
|
|
qui replace ``j'' = `r' if ``j'' == `=`r'+`minm_`j'''
|
|
qui replace ``=`j'+`nbitems''' = `r' if ``=`j'+`nbitems''' == `=`r'+`minm_`j'''
|
|
}
|
|
|
|
// Vérif. Que toutes les modas sont utilisées & concordance entre temps
|
|
forvalues m = 0/`=`nbm_`j''-1' {
|
|
qui count if ``j'' == `m' & `gp' == 0
|
|
local nb_rn1_g0 = r(N)
|
|
qui count if ``j'' == `m' & `gp' == 1
|
|
local nb_rn1_g1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `m' & `gp' == 0
|
|
local nb_rn2_g0 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `m' & `gp' == 1
|
|
local nb_rn2_g1 = r(N)
|
|
local nb_rn = min(`nb_rn1_g0',`nb_rn2_g0',`nb_rn1_g1',`nb_rn2_g1')
|
|
|
|
if `nb_rn' == 0 { // Une moda n'est pas utilisée
|
|
local recoda_`j' = 1
|
|
if `m' == 0 | `m' < `minm`j'_t1_g0' | `m' < `minm`j'_t2_g0' | `m' < `minm`j'_t1_g1' | `m' < `minm`j'_t2_g1' { // La moda 0 n'est pas utilisée
|
|
local stop = 1
|
|
forvalues k = 1/`=`nbm_`j''-`m'' {
|
|
qui count if ``j'' == `=`m' + `k'' & `gp' == 0
|
|
local v`k'1_0 = r(N)
|
|
qui count if ``j'' == `=`m' + `k'' & `gp' == 1
|
|
local v`k'1_1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' + `k'' & `gp' == 0
|
|
local v`k'2_0 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' + `k'' & `gp' == 1
|
|
local v`k'2_1 = r(N)
|
|
if (`v`k'1_0' != 0 | `v`k'2_0' != 0 | `v`k'1_1' != 0 | `v`k'2_1' != 0) & `stop' != 0 {
|
|
qui replace ``j''= `=`m'+`k'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''=`=`m'+`k'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `=`m'+`k'' merged"
|
|
local stop = 0
|
|
}
|
|
}
|
|
}
|
|
else if `m' == `=`nbm_`j''-1' | `m' >= `maxm`j'_t2_g0' | `m' >= `maxm`j'_t1_g1' | `m' >= `maxm`j'_t2_g1' { // La moda max n'est pas utilisée
|
|
local stop = 1
|
|
forvalues k = 1/`=`m'' {
|
|
qui count if ``j'' == `=`m' - `k'' & `gp' == 0
|
|
local v`k'1_0 = r(N)
|
|
qui count if ``j'' == `=`m' - `k'' & `gp' == 1
|
|
local v`k'1_1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' - `k'' & `gp' == 0
|
|
local v`k'2_0 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' - `k'' & `gp' == 1
|
|
local v`k'2_1 = r(N)
|
|
if (`v`k'1_0' != 0 | `v`k'2_0' != 0 | `v`k'1_1' != 0 | `v`k'2_1' != 0 ) & `stop' != 0 {
|
|
qui replace ``j''= `=`m' - `k'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''= `=`m' - `k'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `=`m'-`k'' merged"
|
|
local stop = 0
|
|
}
|
|
}
|
|
}
|
|
else { // Moda central non utilisée
|
|
if runiform()>0.5{ // Tirage au sort pour regrouper
|
|
local stop = 1
|
|
forvalues k = 1/`m' {
|
|
qui count if ``j'' == `=`m' - `k'' & `gp' == 0
|
|
local v`k'1_0 = r(N)
|
|
qui count if ``j'' == `=`m' - `k'' & `gp' == 1
|
|
local v`k'1_1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' - `k'' & `gp' == 0
|
|
local v`k'2_0 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' - `k'' & `gp' == 1
|
|
local v`k'2_1 = r(N)
|
|
if (`v`k'1_0' != 0 | `v`k'2_0' != 0 | `v`k'1_1' != 0 | `v`k'2_1' != 0) & `stop' != 0 {
|
|
qui replace ``j''= `=`m'-`k'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''=`=`m'-`k'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `=`m'-`k'' merged"
|
|
local stop = 0
|
|
}
|
|
}
|
|
}
|
|
else {
|
|
local stop = 1
|
|
forvalues k = 1/`=`nbm_`j''-`m'' {
|
|
qui count if ``j'' == `=`m' + `k'' & `gp' == 0
|
|
local v`k'1_0 = r(N)
|
|
qui count if ``j'' == `=`m' + `k'' & `gp' == 1
|
|
local v`k'1_1 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' + `k'' & `gp' == 0
|
|
local v`k'2_0 = r(N)
|
|
qui count if ``=`j'+`nbitems''' == `=`m' + `k'' & `gp' == 1
|
|
local v`k'2_1 = r(N)
|
|
if (`v`k'1_0' != 0 | `v`k'2_0' != 0 | `v`k'1_1' != 0 | `v`k'2_1' != 0) & `stop' != 0{
|
|
qui replace ``j''=`=`m' + `k'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''=`=`m' + `k'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `=`m'+`k'' merged"
|
|
local stop = 0
|
|
}
|
|
else {
|
|
if `stop' != 0 {
|
|
qui replace ``j''= `nbm_`j'' if ``j''==`m'
|
|
qui replace ``=`j'+`nbitems'''= `nbm_`j'' if ``=`j'+`nbitems'''==`m'
|
|
di "WARNING: items ``j'' & ``=`j'+`nbitems''': answers `m' and `nbm_`j'' merged"
|
|
local stop = 0
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
if `com_z' == 1 {
|
|
di
|
|
di "WARNING : Automatic recoding, the first response category is 0. see {help rosali:help rosali}."
|
|
di
|
|
}
|
|
|
|
forvalues j =1/`nbitems' {
|
|
qui tab ``j'', matrow(rec) // Récupération des infos moda du temps 1
|
|
local nbm`j'_t1 = r(r)
|
|
|
|
qui tab ``=`j'+`nbitems''' // Récupération des infos moda du temps 2
|
|
local nbm`j'_t2 = r(r)
|
|
|
|
local nbm_`j' = max(`nbm`j'_t1', `nbm`j'_t2')
|
|
//Recodage des réponses en 0, 1, 2, etc...
|
|
forvalues r = 0/`=`nbm_`j''-1' {
|
|
qui replace ``j'' = `r' if ``j'' == `=rec[`=`r'+1',1]'
|
|
qui replace ``=`j'+`nbitems''' = `r' if ``=`j'+`nbitems''' == `=rec[`=`r'+1',1]'
|
|
}
|
|
}
|
|
|
|
/* Calcul de nbmoda & nbdif */
|
|
forvalues j = 1/`nbitems' {
|
|
qui tab ``j''
|
|
local nbmoda_`j' = r(r)
|
|
local nbdif_`j' = r(r) - 1
|
|
}
|
|
|
|
local maxdif = 0
|
|
local nbmoda_sum = 0
|
|
forvalues j = 1/`nbitems' {
|
|
if `maxdif' < `nbdif_`j'' {
|
|
local maxdif = `nbdif_`j''
|
|
}
|
|
local nbmoda_sum = `nbmoda_sum' + `nbdif_`j''
|
|
}
|
|
|
|
/* Au moins 2 moda par item */
|
|
forvalues j=1/`nbitems' {
|
|
if `nbmoda_`j'' == 1 {
|
|
di as error "``j'' have only one response category, the analysis can be performed only if each item has at least 2 response categories"
|
|
error 198
|
|
exit
|
|
}
|
|
}
|
|
|
|
local coln ""
|
|
forvalues j =1 /`nbitems' {
|
|
local coln "`coln' ``j''"
|
|
}
|
|
|
|
matrix nbmod = J(2,`nbitems',.)
|
|
|
|
matrix colnames nbmod = `coln'
|
|
matrix rownames nbmod = NbModa Recoding
|
|
|
|
forvalues j = 1/`nbitems' {
|
|
matrix nbmod[1,`j'] = `nbmoda_`j''
|
|
matrix nbmod[2,`j'] = `recoda_`j''
|
|
}
|
|
|
|
*Erreur si plus de 200 difficultés
|
|
local nb_test = 0
|
|
forvalues j=1/`nbitems' {
|
|
local nb_test = `nb_test'+`nbmoda_`j'' -1
|
|
}
|
|
|
|
if `nb_test' >= 200 {
|
|
di as error "The total number of items difficulties to be estimated must be less than 200 ({hi:moda} option option)."
|
|
error 198
|
|
exit
|
|
}
|
|
|
|
local nbitp = 0
|
|
|
|
forvalues j = 1/`nbitems' {
|
|
if `nbmoda_`j'' >= 2 {
|
|
local nbitp = `nbitp' + 1
|
|
}
|
|
}
|
|
|
|
qui count
|
|
local nbpat = r(N)
|
|
|
|
|
|
/*********************************
|
|
* DEFINITION DES CONTRAINTES
|
|
*********************************/
|
|
|
|
if "`group'"!="" { // Contraintes si option groupe
|
|
*EGALITE ENTRE GROUPES A T1 (1-200)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
constraint `=0+`maxdif'*(`j'-1)+`p'' [`p'.``j'']0bn.`gp'=[`p'.``j'']1.`gp'
|
|
}
|
|
}
|
|
|
|
*DIF UNIFORME A T1 (201-400)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=2/`nbdif_`j''{
|
|
constraint `=200+`maxdif'*(`j'-1)+`p'' [`p'.``j'']1.`gp'-[`p'.``j'']0bn.`gp'=`p'*[1.``j'']1.`gp'-`p'*[1.``j'']0bn.`gp'
|
|
}
|
|
}
|
|
|
|
*EGALITES ENTRE T1 et T2, groupe 0 (401-600)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
constraint `=400+`maxdif'*(`j'-1)+`p'' [`p'.``j'']0bn.`gp'=[`p'.``=`j'+`nbitems''']0bn.`gp'
|
|
}
|
|
}
|
|
|
|
*EGALITES ENTRE T1 et T2, groupe 1 (601-800)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
constraint `=600+`maxdif'*(`j'-1)+`p'' [`p'.``j'']1.`gp'=[`p'.``=`j'+`nbitems''']1.`gp'
|
|
}
|
|
}
|
|
|
|
* RC COMMUNE (801-1000)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
constraint `=800+`maxdif'*(`j'-1)+`p'' [`p'.``=`j'+`nbitems''']0bn.`gp'-[`p'.``j'']0bn.`gp'=[`p'.``=`j'+`nbitems''']1.`gp'-[`p'.``j'']1.`gp'
|
|
}
|
|
}
|
|
|
|
* RC UNIFORME, groupe 0 (1001-1200)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=2/`nbdif_`j''{
|
|
constraint `=1000+`maxdif'*(`j'-1)+`p'' `p'*([1.``=`j'+`nbitems''']0bn.`gp'-[1.``j'']0bn.`gp')=[`p'.``=`j'+`nbitems''']0bn.`gp'-[`p'.``j'']0bn.`gp'
|
|
}
|
|
}
|
|
|
|
* RC UNIFORME, groupe 1 (1201-1400)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=2/`nbdif_`j''{
|
|
constraint `=1200+`maxdif'*(`j'-1)+`p'' `p'*([1.``=`j'+`nbitems''']1.`gp'-[1.``j'']1.`gp')=[`p'.``=`j'+`nbitems''']1.`gp'-[`p'.``j'']1.`gp'
|
|
}
|
|
}
|
|
|
|
*Sans interaction temps x groupe
|
|
constraint 1999 [/]:mean(THETA2)#1.`gp'-[/]:mean(THETA2)#0bn.`gp'=[/]:mean(THETA1)#1.`gp'-[/]:mean(THETA1)#0bn.`gp'
|
|
}
|
|
else { //Contraintes si pas d'option groupe
|
|
*EGALITE ENTRE T1 et T2 (401-600)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
constraint `=400+`maxdif'*(`j'-1)+`p'' [`p'.``j'']:_cons = [`p'.``=`j'+`nbitems''']:_cons
|
|
}
|
|
}
|
|
*RC UNIFORME (1001-1200)
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=2/`nbdif_`j''{
|
|
constraint `=1000+`maxdif'*(`j'-1)+`p'' `p'*([1.``=`j'+`nbitems''']:_cons - [1.``j'']:_cons)=[`p'.``=`j'+`nbitems''']:_cons -[`p'.``j'']:_cons
|
|
}
|
|
}
|
|
}
|
|
|
|
/*********************************
|
|
* MATRICE DES RESULTATS
|
|
*********************************/
|
|
matrix dif_rc=J(`nbitems',8,.)
|
|
matrix colnames dif_rc=DIFT1 DIFU RC RC_DIF RCG0 RCUG0 RCG1 RCUG1
|
|
local rown ""
|
|
|
|
forvalues j =1 /`nbitems' {
|
|
local rown "`rown' ``j''"
|
|
}
|
|
matrix rownames dif_rc = `rown'
|
|
|
|
*Nb modalité max
|
|
local nbdif_max = 0
|
|
forvalues j=1/`nbitems' {
|
|
if `nbdif_max' < `nbdif_`j'' {
|
|
local nbdif_max = `nbdif_`j''
|
|
}
|
|
}
|
|
|
|
////////////////////////////////////////////////////////////////////////////////
|
|
////////////////////////////////////////////////////////////////////////////////
|
|
////////////////////////////////////////////////////////////////////////////////
|
|
//////// PARTIE 1 : DIF A T1 ? ////////
|
|
////////////////////////////////////////////////////////////////////////////////
|
|
////////////////////////////////////////////////////////////////////////////////
|
|
////////////////////////////////////////////////////////////////////////////////
|
|
|
|
if "`group'"!="" & "`nodif'"=="" { // PARTIE 1 = Slmt si option group & pas de "nodif"
|
|
|
|
di _dup(49) "_ "
|
|
di
|
|
*di as input "PART 1: DETECTION OF DIFFERENCE IN ITEM DIFFICULTIES BETWEEN GROUPS AT TIME 1"
|
|
|
|
*********************************
|
|
** MODEL B **
|
|
*********************************
|
|
|
|
local model ""
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
local model "`model' (`p'.``j''<-THETA@`p')"
|
|
}
|
|
}
|
|
|
|
qui gsem `model', mlogit tol(0.01) iterate(100) group(`gp') ginvariant(coef loading cons) var(0: THETA@v) var(1:THETA@v) latent(THETA) nocapslatent
|
|
/* Stockage des estimations du modèle */
|
|
estimates store modeldifB
|
|
matrix val_mB = r(table)
|
|
matrix esti_B = e(b)
|
|
|
|
|
|
/* Calcul des difficultés d'item (delta_j) */
|
|
matrix delta_mB=J(`nbitems',`=`nbdif_max'*2',.)
|
|
local name_partOneC ""
|
|
|
|
forvalues p=1/`nbdif_max' {
|
|
forvalues g=0/1 {
|
|
local name_partOneC "`name_partOneC' delta_`p'_gp`g'"
|
|
}
|
|
}
|
|
local name_partOneL ""
|
|
|
|
forvalues j=1/`nbitems' {
|
|
local name_partOneL "`name_partOneL' ``j''"
|
|
}
|
|
|
|
matrix colnames delta_mB = `name_partOneC'
|
|
matrix rownames delta_mB = `name_partOneL'
|
|
matrix delta_mB_se=J(`nbitems',`=`nbdif_max'*2',.)
|
|
local name_partOneC_se ""
|
|
|
|
forvalues p=1/`nbdif_max' {
|
|
forvalues g=0/1 {
|
|
local name_partOneC_se "`name_partOneC_se' delta_`p'_gp`g'_se"
|
|
}
|
|
}
|
|
|
|
matrix colnames delta_mB_se = `name_partOneC_se'
|
|
matrix rownames delta_mB_se = `name_partOneL'
|
|
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
forvalues g=0/1{
|
|
qui lincom -[`p'.``j'']:`g'.`gp'
|
|
local delta`j'_`p'g`g'mB=r(estimate)
|
|
local delta`j'_`p'g`g'mB_se=r(se)
|
|
if `p'>1{
|
|
qui lincom [`=`p'-1'.``j'']:`g'.`gp' - [`p'.``j'']:`g'.`gp'
|
|
local delta`j'_`p'g`g'mB = r(estimate)
|
|
local delta`j'_`p'g`g'mB_se = r(se)
|
|
}
|
|
matrix delta_mB[`j',`=2*`p'-1+`g'']=`delta`j'_`p'g`g'mB'
|
|
matrix delta_mB_se[`j',`=2*`p'-1+`g'']=`delta`j'_`p'g`g'mB_se'
|
|
}
|
|
}
|
|
}
|
|
|
|
matrix var_mB = (val_mB[1,"/var(THETA)#0bn.`gp'"]\val_mB[2,"/var(THETA)#0bn.`gp'"])
|
|
|
|
/*group effect*/
|
|
qui lincom [/]:mean(THETA)#1.`gp'-[/]:mean(THETA)#0bn.`gp'
|
|
local geffmB=r(estimate)
|
|
local segeffmB=r(se)
|
|
qui test [/]:mean(THETA)#1.`gp'-[/]:mean(THETA)#0bn.`gp'=0
|
|
local gcmBp=r(p)
|
|
local gcmBchi=r(chi2)
|
|
local gcmBdf=r(df)
|
|
|
|
|
|
*********************************
|
|
** MODEL A **
|
|
*********************************
|
|
|
|
local model ""
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
local model "`model' (`p'.``j''<-THETA@`p')"
|
|
}
|
|
}
|
|
|
|
qui gsem `model', mlogit tol(0.01) iterate(100) group(`gp') ginvariant(coef loading means) var(0: THETA@v) var(1:THETA@v) from(esti_B, skip) latent(THETA) nocapslatent
|
|
|
|
/* Stockage des estimations du modèle */
|
|
estimates store modeldifA
|
|
matrix val_mA = r(table)
|
|
matrix esti_A = e(b)
|
|
|
|
/* Calcul des difficultés d'item (delta_j) */
|
|
matrix delta_mA=J(`nbitems',`=`nbdif_max'*2',.)
|
|
local name_partOneC ""
|
|
|
|
forvalues p=1/`nbdif_max' {
|
|
forvalues g=0/1 {
|
|
local name_partOneC "`name_partOneC' delta_`p'_gp`g'"
|
|
}
|
|
}
|
|
local name_partOneL ""
|
|
|
|
forvalues j=1/`nbitems' {
|
|
local name_partOneL "`name_partOneL' ``j''"
|
|
}
|
|
matrix colnames delta_mA = `name_partOneC'
|
|
matrix rownames delta_mA = `name_partOneL'
|
|
matrix delta_mA_se=J(`nbitems',`=`nbdif_max'*2',.)
|
|
local name_partOneC_se ""
|
|
|
|
forvalues p=1/`nbdif_max' {
|
|
forvalues g=0/1 {
|
|
local name_partOneC_se "`name_partOneC_se' delta_`p'_gp`g'_se"
|
|
}
|
|
}
|
|
|
|
matrix colnames delta_mA_se = `name_partOneC_se'
|
|
matrix rownames delta_mA_se = `name_partOneL'
|
|
|
|
forvalues j=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`j''{
|
|
forvalues g=0/1{
|
|
qui lincom -[`p'.``j'']:`g'.`gp'
|
|
local delta`j'_`p'g`g'mA=r(estimate)
|
|
local delta`j'_`p'g`g'mA_se=r(se)
|
|
if `p'>1{
|
|
qui lincom [`=`p'-1'.``j'']:`g'.`gp' - [`p'.``j'']:`g'.`gp'
|
|
local delta`j'_`p'g`g'mA = r(estimate)
|
|
local delta`j'_`p'g`g'mA_se = r(se)
|
|
}
|
|
matrix delta_mA[`j',`=2*`p'-1+`g'']=`delta`j'_`p'g`g'mA'
|
|
matrix delta_mA_se[`j',`=2*`p'-1+`g'']=`delta`j'_`p'g`g'mA_se'
|
|
}
|
|
}
|
|
}
|
|
//Variance et se mA
|
|
matrix var_mA = (val_mA[1,"/var(THETA)#0bn.`gp'"]\val_mA[2,"/var(THETA)#0bn.`gp'"])
|
|
|
|
*********************************
|
|
*************MODEL C*************
|
|
*********************************
|
|
// Etape itérative si lrtest significatif
|
|
local nb_stepC = 0
|
|
qui lrtest modeldifA modeldifB
|
|
local diftestp=r(p)
|
|
if `diftestp'<0.05{ /*If pvalue(LRtest)<0.05 then step C*/
|
|
di
|
|
*di as input "PROCESSING STEP C"
|
|
di
|
|
|
|
/*test DIF pour chaque item*/
|
|
local boucle = 1
|
|
local stop = 0
|
|
while `boucle'<=`=`nbitp'-1' & `stop'==0{ /*on s'arrête quand on a libéré du DIF sur (tous les items-1) ou lorsqu'il n'y a plus de tests significatifs*/
|
|
local nb_stepC = `boucle'
|
|
local pajust=0.05
|
|
/*réinitialisation de la matrice de test*/
|
|
matrix test_difu_`boucle'=J(`nbitems',3,.)
|
|
matrix colnames test_difu_`boucle'=chi_DIFU df_DIFU pvalueDIFU
|
|
matrix test_dif_`boucle'=J(`nbitems',3,.)
|
|
matrix colnames test_dif_`boucle'=chi_DIF df_DIF pvalueDIF
|
|
local nbsig=0
|
|
local minpval=1
|
|
local itemdif=0
|
|
if "`detail'" != ""{
|
|
|
|
di as text "Loop `boucle'"
|
|
di as text _col(5) "Adjusted alpha: " %6.4f `pajust'
|
|
di
|
|
di as text _col(10) "{hline 65}"
|
|
di as text _col(10) "Freed item" _col(31) "Chi-Square" _col(48) "DF" _col(57) "P-Value"
|
|
di as text _col(10) "{hline 65}"
|
|
}
|
|
/*boucle de test*/
|
|
forvalues j=1/`nbitems'{
|
|
//if `nbdif_`j'' > 2 {
|
|
local model ""
|
|
local listconst ""
|
|
if dif_rc[`j',1]==. | dif_rc[`j',1]==0 { /*si pas de DIF déjà détecté sur l'item j*/
|
|
/*on libère le DIF de l'item i: pas de contraintes*/
|
|
forvalues k=1/`nbitems'{ /*contraintes pour les autres items (si DIF NU sur item k, pas de contraintes*/
|
|
if `k'!=`j' & `nbmoda_`j'' >= 2 {
|
|
if dif_rc[`k',1]==. | dif_rc[`k',1]==0 {/*pas de DIF sur item k: contraintes 1-200*/
|
|
forvalues p=1/`nbdif_`k''{
|
|
qui local listconst "`listconst' `=0+`maxdif'*(`k'-1)+`p''"
|
|
qui constraint list `=0+`maxdif'*(`k'-1)+`p''
|
|
}
|
|
}
|
|
else{
|
|
if dif_rc[`k',2]!=. & dif_rc[`k',2]!= 0 & `nbmoda_`k'' > 2 { /*DIF U: contraintes 201-400*/
|
|
forvalues p=2/`nbdif_`k''{
|
|
qui local listconst "`listconst' `=200+`maxdif'*(`k'-1)+`p''"
|
|
qui constraint list `=200+`maxdif'*(`k'-1)+`p''
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
forvalues jj=1/`nbitems'{
|
|
forvalues p=1/`nbdif_`jj''{
|
|
local model "`model' (`p'.``jj''<-THETA@`p')"
|
|
}
|
|
}
|
|
qui gsem `model', mlogit tol(0.01) iterate(100) group(`gp') ginvariant(coef loading) var(0: THETA@v) var(1:THETA@v) constraint(`listconst') from(esti_B) latent(THETA) nocapslatent
|
|
estimates store modeldif3b`boucle'it`i'
|
|
|
|
*************************
|
|
*****test DIF item i*****
|
|
*************************
|
|
qui test [1.``j'']1.`gp'=[1.``j'']0bn.`gp'
|
|
if `nbmoda_`j'' > 2 {
|
|
forvalues p=2/`nbdif_`j''{
|
|
qui test [`p'.``j'']1.`gp'=[`p'.``j'']0bn.`gp', acc
|
|
}
|
|
}
|
|
matrix test_dif_`boucle'[`j',1]=(r(chi2),r(df),r(p))
|
|
|
|
/* Test DIF Uniforme */
|
|
if `nbmoda_`j'' > 2 {
|
|
qui test 2*([1.``j'']1.`gp'-[1.``j'']0bn.`gp')=[2.``j'']1.`gp'-[2.``j'']0bn.`gp'
|
|
forvalues p=3/`nbdif_`j''{
|
|
qui test `p'*([1.``j'']1.`gp'-[1.``j'']0bn.`gp')=[`p'.``j'']1.`gp'-[`p'.``j'']0bn.`gp', acc
|
|
}
|
|
matrix test_difu_`boucle'[`j',1]=(r(chi2), r(df), r(p))
|
|
}
|
|
|
|
if test_dif_`boucle'[`j',3]<`pajust'{/*si DIF sur item i*/
|
|
local ++nbsig
|
|
if test_dif_`boucle'[`j',3]<`minpval'{
|
|
local minpval=test_dif_`boucle'[`j',3]
|
|
local itemdif=`j'
|
|
}
|
|
}
|
|
if "`detail'" != "" {
|
|
di as text _col(10) abbrev("``j''",15) as result _col(31) %6.3f test_dif_`boucle'[`j',1] _col(48) test_dif_`boucle'[`j',2] _col(57) %6.4f test_dif_`boucle'[`j',3]
|
|
}
|
|
}
|
|
}
|
|
/*si nb de tests significatifs=0, on arrête*/
|
|
if `nbsig'==0{
|
|
local stop=1
|
|
if `boucle' == 1 {
|
|
if "`detail'" != "" {
|
|
di as text _col(10) "{hline 65}"
|
|
di
|
|
di as result "No significant test: no difference between groups detected, no DIF detected"
|
|
di
|
|
}
|
|
}
|
|
else {
|
|
if "`detail'" != ""{
|
|
di as text _col(10) "{hline 65}"
|
|
di
|
|
di as result "No other significant tests"
|
|
di
|
|
}
|
|
}
|
|
}
|
|
else{/*si nb de tests significatifs>0, mise à jour de la matrice de résultats*/
|
|
matrix dif_rc[`itemdif',1]=`boucle'
|
|
if "`detail'" != ""{
|
|
di as text _col(10) "{hline 65}"
|
|
di
|
|
di as result "Difference between groups on ``itemdif'' at time 1"
|
|
}
|
|
if `nbmoda_`itemdif'' > 2 {
|
|
if "`detail'" != "" {
|
|
|
|
di
|
|
di %~60s as text "Test of uniform difference"
|
|
di _col(10) "{hline 40}"
|
|
di _col(10) as text "Chi-square" _col(28) "DF" _col(40) "P-value"
|
|
di _col(10) as result %4.2f `=test_difu_`boucle'[`itemdif',1]' _col(28) `=test_difu_`boucle'[`itemdif',2]' _col(40) %4.2f `=test_difu_`boucle'[`itemdif',3]'
|
|
di _col(10) as text "{hline 40}"
|
|
}
|
|
if test_difu_`boucle'[`itemdif',3]<0.05{ /*DIF NU détectée*/
|
|
matrix dif_rc[`itemdif',2]=0
|
|
di
|
|
di as result "``itemdif'' : Non-uniform differences of item difficulties between groups at T1"
|
|
di
|
|
}
|
|
else{/*DIF U détectée*/
|
|
matrix dif_rc[`itemdif',2]=`boucle'
|
|
di
|
|
di as result "``itemdif'' : Uniform differences of item difficulties between groups at T1"
|
|
di
|
|
}
|
|
}
|
|
else {
|
|
// Différence entre groupes au temps 1 mais slmt 2 moda. donc pas de U ou NU
|
|
di _col(15) _dup(60) "-"
|
|
}
|
|
}
|
|
local ++boucle
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
matrix dif_detect = J(1,2*`nbitems',.)
|
|
local numdif=1
|
|
forvalues j=1/`nbitems' {
|
|
if dif_rc[`j',1] != . {
|
|
matrix dif_detect[1,`numdif']=`j'
|
|
if dif_rc[`j',2] == 0 {
|
|
matrix dif_detect[1,`nbitems'+`numdif']=0
|
|
}
|
|
if dif_rc[`j',2] != 0 {
|
|
matrix dif_detect[1,`nbitems'+`numdif']=1
|
|
}
|
|
local numdif = `numdif'+1
|
|
}
|
|
}
|
|
return matrix difitems = dif_detect
|
|
|
|
capture qui use `saverspcm', clear
|
|
|
|
end
|