* This example file shows how you can set starting values
* for what observations get dropped. 


clear
infile y x z using smpldata.raw

local depvar y    /* Dependent Variable */
local expvar x z  /* Explanatory Variables */
local quint 0.35  /* Quantile of Interest */


*qcenreg `depvar' `expvar', quantile(`quint')

qreg `depvar' `expvar', quantile(0.45)

predict predval
egen lowval = min(`depvar')
qcenreg `depvar' `expvar' if predval > lowval, quantile(`quint')


qcrstep `depvar' `expvar', quantile(`quint') qstart(0.45) qstep(0.02)




* to get bootstrapped standard errors


local nreps 100      /* Number of replications */
local dname "bsest"  /* Name of the dataset to store estimates */
local niters 40      /* Max number of iterations to try to improve the quantile estimate*/

bs " qcenreg `depvar' `expvar', quantile(`quint')  maxiter(`niters')" "_b[x] _b[z] _b[_cons]", reps(`nreps') saving(`dname') replace dots


bs " qcrstep `depvar' `expvar', quantile(`quint') qstart(0.45) qstep(0.02) maxiter(`niters')" "_b[x] _b[z] _b[_cons]", reps(`nreps') saving(`dname') replace dots


* reps controls how many replications to do
* saving where to put the estimates (optional but I think useful)
* replace says to replace the dataset if it already exists.
* dots puts a dot on the screen for each replication