[George Hoffman]

xtqreg.ado (* Version 1.0 - 30 Dec 2017) has no reference to 'r.exe'. is your Stata installation correct? perhaps you have conflicts in adopath or another user-installed xtqreg.ado?

[Jon Burzka]

thanks. those are the installations i did so far. dont know how I could check for such issues.


[1] package fitstat from http://fmwww.bc.edu/repec/bocode/f
'FITSTAT': module to compute fit statistics for single equation regression mode
> ls

[2] package grqreg from http://fmwww.bc.edu/repec/bocode/g
'GRQREG': module to graph the coefficients of a quantile regression

[3] package outreg from http://fmwww.bc.edu/repec/bocode/o
'OUTREG': module to write estimation tables to a Word or TeX file

[4] package xtqreg from http://www.imm.ki.se/biostatistics/stata
'XTQREG': Linear Quantile Mixed Models

[5] package rsource from http://fmwww.bc.edu/repec/bocode/r
'RSOURCE': module to run R from inside Stata using an R source file

[6] package qregpd from http://fmwww.bc.edu/repec/bocode/q
'QREGPD': module to perform Quantile Regression for Panel Data

[7] package genqreg from http://fmwww.bc.edu/repec/bocode/g
'GENQREG': module to perform Generalized Quantile Regression

[8] package amcmc from http://fmwww.bc.edu/repec/bocode/a
'AMCMC': module to provide Mata functions and structures for adaptive Markov ch
> ain Monte Carlo sampling

[9] package moremata from http://fmwww.bc.edu/repec/bocode/m
'MOREMATA': module (Mata) to provide various functions

[10] package rif from http://fmwww.bc.edu/repec/bocode/r

[Joao Santos Silva]

On [4] I see that you installed a different package called xtqreg (I was not aweare of it); please install xtqreg from ssc using the command "ssc install qreg, replace"

Best wishes,

Joao

[Jon Burzka]

thanks, it works! great!

the significance levels are way lower.

Now I still need to figure: WARNING: some fitted values of the scale function are negative.
How can I shift to fixed effects on individual basis only as now my time fix variables are out?
Do those fixed effects now actually account for country or country-pair fixed effects?

[Joao Santos Silva]

The warning means that you probably should change the specification of your model.
The fixed effects are for the groups defined by xtset

[Jon Burzka]

great! thanks again. interesting..
so I should set "xtset year" to recieve individual fixed effects without time fixed effects?
(the static variables appear now)

so do I have possibilities to insert robust standard errors?

[Joao Santos Silva]

if you set "xtset year" you will get year FEs.

[Jon Burzka]

I am confused by that. Why do my constant over time variables drop out when I set "xtset individual"? I mean these variables change by individual but not over time.
As far as I know time fixed effects cover the change of variables over time, so if I set "xtset year" my constant over time variables of interest should drop.

[Joao Santos Silva]

Time-constant variables will drop if you xtset by individual, if you xtset by year the variables that droped are the ones that are constant across individuals.

[Jon Burzka]

thanks! I had a look, I mixed it up.. Now only one gap remains. Can I use boostrapped SE?

Best
Burz

[Joao Santos Silva]

You can bootstrap the command.

[Jon Burzka]

thanks for your reply. As vce of course doesn't work, could you tell me how to syntax?

Appart from that. If I use "sqreg y x i.id" this creates dummies for each country but delivers completely absurd coefficients.. Shouldn't this deliver similar results than xtqreg?

Best
Burz

[Joao Santos Silva]

Please check the help file for the bootstrap command. About sqreg, I would expect it to give different and unreliable results; that is why we wrote a new paper and a new command :-)

[Jon Burzka]

ah, this makes sense of course, now the circle closes. Thanks.
So I figured the following:
bootstrap, reps(200) seed(20): xtqreg $ylist $xlist, quantile(.1 .25 .5 .75)


varlist required
r(100);

I assume that a further specification is missing? I specified the data and variables up front so this should be fine.

[Joao Santos Silva]

There is a simple workaround for that, but I am about to release a new version of the command that addresses the problem (and allows for factor variables). If you email me, I'll be happy to send it to you.

Best wishes,

Joao