//Author: @Niranjan Kumar, CAFRAL
//regression post-estimation application; quite useful in counter-factual analysis
//First Application: Bank Entry, New Loans, and Misallocation by Dr. Nirvana Mitra and Dr. Pavel Chakraborty
cap prog drop coefout
program define coefout
version 10.0 //minimum stata version required for the stata-command to work
syntax varlist(min = 1), [pmin(string) pmax(string)] [labelname(string)] [keep(string)]
//save the coefficient vector
mat coef = e(b)
//save the variance-covariance vector
mat cov = e(V)
//residual degrees of freedom
scalar df= e(df_r)
*****************************************************************************
//first cut: syntax checkers
*****************************************************************************
//check if both varlist and labelname are not empty
if "`varlist'" == "" & "`labelname'" == ""{
di in bl "Nothing to do here since option varname and variable list both are empty"
exit 0
}
local keepval = ""
//check the option keep and create a default version for it
if "`keep'" ==""{
local keepval= "def"
}
else{
local keepval: di "`keep'"
}
*****************************************************************************
// part-1 : calculate standard-errors, t-stats, p-values
*****************************************************************************
//variance-covariance matrix : number of rows
local n = rowsof(cov)
//row and column size of coeficient vector
local num_rows1 = rowsof(coef) //It is 1 but in case if multiple regressions saved for same variable names
local num_cols1= colsof(coef)
//create a matrix to store the diagonal elements
matrix cov_diag = J(`n', 1, .)
//loop over the diagonal elements and copy them to the se_diag matrix
forval m = 1/`n' {
matrix cov_diag[`m', 1] = cov[`m', `m']
}
//a blank standard error vector generated
matrix se= J(`n', 1, .)
// take a square root of the covariance-variance diagonal vector
forval m = 1/`n' {
matrix se[`m', 1] = sqrt(cov_diag[`m', 1])
}
//number of rows and columns of standard error vector
matrix coef_new = coef'
//compute the t-stats
matrix t = J(`n', 1, .)
forval m = 1/`n'{
if se[`m', 1] != . & coef_new[`m', 1] !=. {
matrix t[`m', 1] = coef_new[`m', 1]/se[`m', 1]
}
else {
matrix t[`m', 1] = 0.0
}
}
//compute the p-value
matrix p = J(`n', 1, .)
forval m = 1/`n' {
scalar t_stat = t[`m', 1]
scalar p_val = 2*ttail(df, abs(t_stat))
matrix p[`m', 1] = p_val
}
matrix p1 = p
global num_rows1 : rowfullnames coef_new
matrix rownames p1 = $num_rows1
matrix rownames se = $num_rows1
matrix rownames t = $num_rows1
********************************************************************************
//part-2 : syntax analysis
********************************************************************************
//option pmin and pmax syntax check:
if "`pmin'" != "" & "`pmax'" == ""{
di in r "Define the range for p-values. pmax value is missing. "
exit 498
}
else if "`pmin'" == "" & "`pmax'" != ""{
di in r "Define the range for p-values. pmin value is missing. "
exit 498
}
local pmin_val = ""
local pmax_val = ""
// handle pmin
if "`pmin'" != ""{
local pmin_val: di "`pmin'"
}
else if "`pmin'" == "" {
local pmin_val = "default_pmin"
}
//handle pmax
if "`pmax'" != ""{
local pmax_val: di "`pmax'"
}
else if "`pmax'" == ""{
local pmax_val = "default_pmax"
}
//handle varlist
local varlist1 = ""
if "`varlist'" != ""{
local varlist1: di "`varlist'"
scalar var_count = `:word count `varlist1''
}
else {
scalar var_count = 0.0
}
//handle labelname
local varnamelist = ""
if "`labelname'" != ""{
local varnamelist: di "`labelname'"
scalar varname_count = `:word count `varnamelist''
}
else{
scalar varname_count = 0.0
}
//total number of variables (varlist and varnames-interaction terms)
scalar total_count = varname_count + var_count
//combine the total variables and variable labels and save them in a local
if "`varlist1'" != "" & "`varnamelist'" != "" {
local final_varlist = "`varlist1'" + " " + "`varnamelist'"
}
else if "`varlist1'" != "" & "`varnamelist'" == ""{
local final_varlist = "`varlist1'"
}
else {
local final_varlist = "`varnamelist'"
}
********************************************************************************
//part-3 output and error handling
//It also checks if any variable name or variable label is incorrect or not matched
********************************************************************************
//case-1.
if strpos("`keepval'", "p")>0 & strpos("`keepval'", "t")==0 & strpos("`keepval'", "se")==0 {
//result matrix; initialized as blank
mat A= J(total_count, 2, .)
scalar shift = 1
foreach var of local final_varlist{
if strpos("$num_rows1", "`var'")>0 {
if strpos("`pmin_val'", "default_pmin")>0 & strpos("`pmax_val'", "default_pmax")>0 {
mat A[shift, 1] = coef_new["`var'", 1]
mat A[shift, 2] = p1["`var'", 1]
}
else {
//filter based on range of p-value
if (p1["`var'", 1] >= `pmin_val' & p1["`var'", 1] <= `pmax_val'){
mat A[shift, 1] = coef_new["`var'", 1]
}
mat A[shift, 2] = p1["`var'", 1]
}
}
else{
di in r "Specified variable name or variable label are not correct."
exit 498
}
scalar shift = shift+ 1
}
mat colnames A = coef p
mat rownames A = `final_varlist'
}
//case-2
else if strpos("`keepval'", "t")>0 & strpos("`keepval'", "p")==0 & strpos("`keepval'", "se")==0 {
//result matrix; initialized as blank
mat A= J(total_count, 2, .)
scalar shift = 1
foreach var of local final_varlist{
if strpos("$num_rows1", "`var'")>0 {
if strpos("`pmin_val'", "default_pmin")>0 & strpos("`pmax_val'", "default_pmax")>0 {
mat A[shift, 1] = coef_new["`var'", 1]
mat A[shift, 2] = t["`var'", 1]
}
else {
//filter based on range of p-value
if (p1["`var'", 1] >= `pmin_val' & p1["`var'", 1] <= `pmax_val'){
mat A[shift, 1] = coef_new["`var'", 1]
}
mat A[shift, 2] = t["`var'", 1]
}
}
else{
di in r "Specified variable name or variable label are not correct."
exit 498
}
scalar shift = shift+ 1
}
mat colnames A = coef t
mat rownames A = `final_varlist'
}
//case-3
else if strpos("`keepval'", "se")>0 & strpos("`keepval'", "t")==0 & strpos("`keepval'", "p")==0 {
//result matrix; initialized as blank
mat A= J(total_count, 2, .)
scalar shift = 1
foreach var of local final_varlist{
if strpos("$num_rows1", "`var'")>0 {
if strpos("`pmin_val'", "default_pmin")>0 & strpos("`pmax_val'", "default_pmax")>0 {
mat A[shift, 1] = coef_new["`var'", 1]
mat A[shift, 2] = se["`var'", 1]
}
else {
//filter based on range of p-value
if (p1["`var'", 1] >= `pmin_val' & p1["`var'", 1] <= `pmax_val'){
mat A[shift, 1] = coef_new["`var'", 1]
}
mat A[shift, 2] = se["`var'", 1]
}
}
else{
di in r "Specified variable name or variable label are not correct."
exit 498
}
scalar shift = shift+ 1
}
mat colnames A = coef se
mat rownames A = `final_varlist'
}
//case-4
else if (strpos("`keepval'", "p t")> 0 | strpos("`keepval'", "t p")>0 ) & (strpos("`keepval'", "se t p")==0 & strpos("`keepval'", "t se p")==0 & strpos("`keepval'", "se p t")==0 & strpos("`keepval'", "p se t")==0 & strpos("`keepval'", "p t se")==0 & strpos("`keepval'", "t p se")==0){
//result matrix; initialized as blank
mat A= J(total_count, 3, .)
scalar shift = 1
foreach var of local final_varlist{
if strpos("$num_rows1", "`var'")>0 {
if strpos("`pmin_val'", "default_pmin")>0 & strpos("`pmax_val'", "default_pmax")>0 {
mat A[shift, 1] = coef_new["`var'", 1]
mat A[shift, 2] = t["`var'", 1]
mat A[shift, 3] = p1["`var'", 1]
}
else {
//filter based on range of p-value
if (p1["`var'", 1] >= `pmin_val' & p1["`var'", 1] <= `pmax_val'){
mat A[shift, 1] = coef_new["`var'", 1]
}
mat A[shift, 2] = t["`var'", 1]
mat A[shift, 3] = p1["`var'", 1]
}
}
else{
di in r "Specified variable name or variable label are not correct."
exit 498
}
scalar shift = shift+ 1
}
mat colnames A = coef t p
mat rownames A = `final_varlist'
}
//case-5
else if (strpos("`keepval'", "p se")> 0 | strpos("`keepval'", "se p")>0) & (strpos("`keepval'", "se t p")==0 & strpos("`keepval'", "t se p")==0 & strpos("`keepval'", "se p t")==0 & strpos("`keepval'", "p se t")==0 & strpos("`keepval'", "p t se")==0 & strpos("`keepval'", "t p se")==0){
//result matrix; initialized as blank
mat A= J(total_count, 3, .)
scalar shift = 1
foreach var of local final_varlist{
if strpos("$num_rows1", "`var'")>0 {
if strpos("`pmin_val'", "default_pmin")>0 & strpos("`pmax_val'", "default_pmax")>0 {
mat A[shift, 1] = coef_new["`var'", 1]
mat A[shift, 2] = se["`var'", 1]
mat A[shift, 3] = p1["`var'", 1]
}
else {
//filter based on range of p-value
if (p1["`var'", 1] >= `pmin_val' & p1["`var'", 1] <= `pmax_val'){
mat A[shift, 1] = coef_new["`var'", 1]
}
mat A[shift, 2] = se["`var'", 1]
mat A[shift, 3] = p1["`var'", 1]
}
}
else{
di in r "Specified variable name or variable label are not correct."
exit 498
}
scalar shift = shift+ 1
}
mat colnames A = coef se p
mat rownames A = `final_varlist'
}
//case-6
else if (strpos("`keepval'", "t se")> 0 | strpos("`keepval'", "se t")>0) & (strpos("`keepval'", "se t p")==0 & strpos("`keepval'", "t se p")==0 & strpos("`keepval'", "se p t")==0 & strpos("`keepval'", "p se t")==0 & strpos("`keepval'", "p t se")==0 & strpos("`keepval'", "t p se")==0){
//result matrix; initialized as blank
mat A= J(total_count, 3, .)
scalar shift = 1
foreach var of local final_varlist{
if strpos("$num_rows1", "`var'")>0 {
if strpos("`pmin_val'", "default_pmin")>0 & strpos("`pmax_val'", "default_pmax")>0 {
mat A[shift, 1] = coef_new["`var'", 1]
mat A[shift, 2] = se["`var'", 1]
mat A[shift, 3] = t["`var'", 1]
}
else {
//filter based on range of p-value
if (p1["`var'", 1] >= `pmin_val' & p1["`var'", 1] <= `pmax_val'){
mat A[shift, 1] = coef_new["`var'", 1]
}
mat A[shift, 2] = se["`var'", 1]
mat A[shift, 3] = t["`var'", 1]
}
}
else{
di in r "Specified variable name or variable label are not correct."
exit 498
}
scalar shift = shift+ 1
}
mat colnames A = coef se t
mat rownames A = `final_varlist'
}
//case-7
else if strpos("`keepval'", "p t se")> 0 | strpos("`keepval'", "p se t")>0 | strpos("`keepval'", "t se p")>0 | strpos("`keepval'", "t p se")>0 |strpos("`keepval'", "se t p")>0 | strpos("`keepval'", "se p t")>0 {
//result matrix; initialized as blank
mat A= J(total_count, 4, .)
scalar shift = 1
foreach var of local final_varlist{
if strpos("$num_rows1", "`var'")>0 {
if strpos("`pmin_val'", "default_pmin")>0 & strpos("`pmax_val'", "default_pmax")>0 {
mat A[shift, 1] = coef_new["`var'", 1]
mat A[shift, 2] = se["`var'", 1]
mat A[shift, 3] = t["`var'", 1]
mat A[shift, 4] = p1["`var'", 1]
}
else {
//filter based on range of p-value
if (p1["`var'", 1] >= `pmin_val' & p1["`var'", 1] <= `pmax_val'){
mat A[shift, 1] = coef_new["`var'", 1]
}
mat A[shift, 2] = se["`var'", 1]
mat A[shift, 3] = t["`var'", 1]
mat A[shift, 4] = p1["`var'", 1]
}
}
else{
di in r "Specified variable name or variable label are not correct."
exit 498
}
scalar shift = shift+ 1
}
mat colnames A = coef se t p
mat rownames A = `final_varlist'
}
//case-8 default
else if strpos("`keepval'", "def")>0 {
//result matrix; initialized as blank
mat A= J(total_count, 1, .)
scalar shift = 1
foreach var of local final_varlist{
if strpos("$num_rows1", "`var'")>0 {
if strpos("`pmin_val'", "default_pmin")>0 & strpos("`pmax_val'", "default_pmax")>0 {
mat A[shift, 1] = coef_new["`var'", 1]
}
else {
//filter based on range of p-value
if (p1["`var'", 1] >= `pmin_val' & p1["`var'", 1] <= `pmax_val'){
mat A[shift, 1] = coef_new["`var'", 1]
}
}
}
else{
di in r "Specified variable name or variable label are not correct."
exit 498
}
scalar shift = shift + 1
}
mat colnames A = coef
mat rownames A = `final_varlist'
}
//case-9: In case if there is anything else in the specification
else{
di in r "Incorrect specification"
exit 498
}
matrix A = A
matlist A
end