qregpd application explanation

[Barbora Sedova]

Dear all,

I am attempting to estimate the effect of temperature and precipitation (at the district level, at t) and their interactions with historical means (HT, Hp respectively) on different quintiles of households’ consumption, using hh fixed effects. I am interested in applying quantile regression for panel data with nonadditive fixed effects (Powell, 2016), using the stata command qregpd.

I attempted to run the following command
qregpd C T TxH P PxH if URBAN==0, id(IDHH) fix(round) optimize(mcmc) noisy draws(1000) burn(100) arate(.5) q(0.15)

where, for demonstration, T refers to temperature and TxH is the interaction term.
I am a bit puzzeled, because all of the estimated coefficients are significant at 1% level. I think there must be something wrong.
I think I am not using the command ina right way.
Even though I read the help file corresponding to the command, I am still not sure what precisely the following part of the command means: draws(1000) burn(100) arate(.5) mean.
Can somebody advice how to practically apply the command?


Thanks so much.

[Scott Merryman]

You might want to read the introduction of the Stata Bayesian Analysis Reference Manual (or other introductory sources). The reference manual also has a glossary with many of these terms.

Burn-in refers to the practice of discarding an initial portion of a Markov chain sample so that the effect of initial values on the posterior inference is minimized. There initial values might land you in a low probability space before it converges. Including these values could basis the estimates.

From the glossary

acceptance rate. In the context of the MH algorithm, acceptance rate is the fraction of the proposed
samples that is accepted. The optimal acceptance rate depends on the properties of the target
distribution and is not known in general. If the target distribution is normal, however, the optimal
acceptance rate is known to be 0.44 for univariate distributions and 0.234 for multivariate
distributions.

As to applying the command, the help file references:

Powell, David. 2014b. Did the Economic Stimulus Payments of 2008 Reduce Labor Supply? Evidence from
Quantile Panel Data Estimation. RAND Labor and Population Working Paper 710-3.

https://www.rand.org/pubs/working_papers/WR710-3.html