wlsiter option in qreg

[haiyan lin]

Dear all,

I was trying to understand what wlsiter do in the qreg regression. When I add many dummies in order to control group fixed effects, there exist convergence problems. The convergence issue disappears when the wlsiter option was added. I wonder what did this option do to solve the convergence problem and what is the consequence of this option? My code works well with wlsiter (100). Would this cause some problems with the estimates?

Hope to hear your suggestion. Thank you in advance!

Best,
Haiyan

[Joro Kolev]

If convergence has been achieved, there is no problem with the estimates. I do not think you should worry about this.

[Joao Santos Silva]

Dear haiyan lin,

I agree with Joro's comment above, but if you want to know more, the help file describes exactly what the wlsiter option does. Also, please note that the qreg command is not particularly suited to deal with models for panel data. Please check the xrqreg command that is available from SSC and check its help file and the references therein.

Best wishes,

Joao

[Chen Samulsion]

wlsiter is abbreviation of "weighted least-squares iteration". -qreg- help file explains:

wlsiter(#) specifies the number of weighted least-squares iterations that will be attempted before the linear programming iterations are started. The default value is 1. If there are convergence problems, increasing this number should help.

see https://www.statalist.org/forums/forum/general-stata-discussion/general/269345-convergence-not-achieved-with-quantile-regression-qreg, especially #4

I would start with just

qreg pcg_faminc , wlsiter(100)

i.e. use qreg to compute the median. That is obviously overkill, but it allows you to see whether there is a problem in the dependent variable, by comparing the number of observations reported by qreg and the number of observations you think it should be, and see if there is something weird about the median.

I would then inspect the month indicator (dummy) variables, by using the user written fre command (which you can install by typing in Stata ssc install fre):

fre $monthdummies

I would look at: are the indicator variables correctly coded (0 or 1), are there any months in which there is only 1 observation, what about missing values.

[haiyan lin]

Thank you so much! All your timely responses are much appreciated! Joro Kolev Joao Santos Silva Chen Samulsion