Quantile on quantile regression model is gaining popularity in providing distribution-based estimates between two variables. But most of the studies have applied quantile on quantile using quantile regression in long timeseries data without accounting for the non stationarity for the variables. This blog provides tutorial for the model using Quantile ARDL which makes model robust to autocorrelation.
View the tutorial at:
Codes are as under
clear
use “C:\Users\LENOVO\Desktop\qantile PMG model\quantile on quantile on stata\data.dta”
tsset time
* Generate scalar values from 5 to 95 with a jump of 5 for quantile values
capture xtile mal_group = mal, nq(100)
gen dpak = d.pak
gen lpak = l.pak
gen dmal = d.mal
gen lmal = l.mal
* QARDL model
qreg dpak lpak dmal lmal
capture postutil clear
tempfile holding
postfile handle qqrji adjji using `holding’
forvalues i = 1/19 {
local tau_indep = `i’ *5
forvalues j = 1/19 {
local tau_val = `j’*5
qreg dpak lpak dmal lmal if mal_group < `tau_indep’, quantile(`tau_val’)
local effect = _b[lmal]
local adj = _b[lpak]
display “Quantile value of Indep. var: ” `tau_indep’ ” & dep var: ” `tau_val’ ” have value: ” `effect’
post handle (`effect’) (`adj’)
}
}
postclose handle
use `holding’, clear
generate PercentileX = ceil(_n / 19) *5
egen obs_in_group = seq(), from(1) to(19)
gen PercentileY = obs_in_group * 5
gen Multiplier = qqrji
gen lrmal = Multiplier / adjji
twoway (contour Multiplier PercentileY PercentileX)
twoway (contour adjji PercentileY PercentileX)
twoway (contour lrmal PercentileY PercentileX)