Dear All,

I have a household-level dataset of 511,877 observations, and run a quantile regression of the log of monthly consumption expenditure (ln_rmpce) on local temperature and precipitation, interacted by NIC_Code (agriculture, industry, services), at the 0.1 quantile. In Stata 17, the convergence fails (with all options, including bsqreg). But, in Stata 15, qreg converges (although the iteration log mentions that alternate solutions exist). The standard errors and pseudo-R2 are reasonable. Kindly guide me on this error, as I need to be able to run this regresion in Stata 17. Thank you.

__________________________________________________ _____________________________________________

Output in Stata/MP 17.0

qreg ln_rmpce c.std_temp#i.NIC_Code c.std_pre#i.NIC_Code i.year i.District_Code_Cruts,
> quantile(.1) vce(robust) wlsiter(999)
Iteration 1: WLS sum of weighted deviations = 93962.545
Iteration 2: WLS sum of weighted deviations = 83182.67
Iteration 3: WLS sum of weighted deviations = 75069.533
Iteration 4: WLS sum of weighted deviations = 68325.169
Iteration 5: WLS sum of weighted deviations = 63227.737
Iteration 6: WLS sum of weighted deviations = 59025.529
.
.
.
Iteration 92: sum of abs. weighted deviations = 41557.604
Iteration 93: sum of abs. weighted deviations = 41557.604
Iteration 94: sum of abs. weighted deviations = 41557.604
VCE computation failed; try increasing the maximum number of iterations or try bsqreg
r(498);

__________________________________________________ ___________________

Output in Stata/IC 15.0

qreg ln_rmpce c.std_temp#i.NIC_Code c.std_pre#i.NIC_Code i.year i.District_Code_Cruts,
> quantile(.1) vce(robust)
.
.
.
Iteration 3606: sum of abs. weighted deviations = 41557.604
Iteration 3607: sum of abs. weighted deviations = 41557.604

.1 Quantile regression Number of obs = 511,877
Raw sum of deviations 208449 (about .75148922)
Min sum of deviations 41557.6 Pseudo R2 = 0.8006

-------------------------------------------------------------------------------------
| Robust
ln_rmpce | Coef. Std. Err. t P>|t| [95% Conf. Interval]
--------------------+----------------------------------------------------------------
NIC_Code#c.std_temp |
1 | .0103571 .0019892 5.21 0.000 .0064582 .0142559
2 | -.0144272 .0026826 -5.38 0.000 -.019685 -.0091694
3 | .0589867 .0020763 28.41 0.000 .0549173 .0630561
|
NIC_Code#c.std_pre |
1 | -.0000569 .0021554 -0.03 0.979 -.0042815 .0041676
2 | .0025723 .002927 0.88 0.380 -.0031646 .0083091
3 | .002321 .0021493 1.08 0.280 -.0018916 .0065336
|
year |
2000 | -4.304993 .0026613 -1617.62 0.000 -4.310209 -4.299777
2005 | .1441158 .0026399 54.59 0.000 .1389417 .1492898
2010 | .203536 .0029251 69.58 0.000 .197803 .2092691
2012 | .354323 .0029233 121.20 0.000 .3485934 .3600527