Hi Statalist!
First of all, thank you for all the help you give on a daily basis! It's been very helpful to lurk here, but now I find myself in a situation where I simply need to ask.
I'm trying to regress Return (depvar) on RMRF SMB HML LIQ (indepvars) using panel data. My panel data set is companynum and date (monthly data 2007-03 to 2016-12), however, it is unbalanced.
Now, when I use the different xtreg options (fe, re, et cetera) and xtgls, I get different results than when I use statsby or regressby. Furthermore, my regression results are quite different than expected, but less so when using statsby or regressby.
I'm now asking: 1) which method to use (xt or statsby), and 2) if I'm using them incorrectly?
I'm posting a smaller sample of my data and my regression commands. Not sure if the dataex output has been formatting correctly for your use, but the columns are date, companynum, RMRF, SMB, HML, LIQ. Date was reformatted in dataex, but is normally in a YYYYmm format (e.g. 2007m3).
The different commands I've been using:
. xtreg Return RMRF SMB HML LIQ, re vce(robust)
statsby _b, by(companynum): reg Return RMRF SMB HML LIQ, robust noconstant
The means of my coefficients from statsby are larger than the coefficients from xtreg. Shouldn't the mean of the coefficients from statsby be equal to the coefficients in the other reg commands, or is it wrong to think of it that way?
Please let me know if you find anything that is incorrectly posted and I will correct accordingly.
Many thanks!
First of all, thank you for all the help you give on a daily basis! It's been very helpful to lurk here, but now I find myself in a situation where I simply need to ask.
I'm trying to regress Return (depvar) on RMRF SMB HML LIQ (indepvars) using panel data. My panel data set is companynum and date (monthly data 2007-03 to 2016-12), however, it is unbalanced.
Now, when I use the different xtreg options (fe, re, et cetera) and xtgls, I get different results than when I use statsby or regressby. Furthermore, my regression results are quite different than expected, but less so when using statsby or regressby.
I'm now asking: 1) which method to use (xt or statsby), and 2) if I'm using them incorrectly?
I'm posting a smaller sample of my data and my regression commands. Not sure if the dataex output has been formatting correctly for your use, but the columns are date, companynum, RMRF, SMB, HML, LIQ. Date was reformatted in dataex, but is normally in a YYYYmm format (e.g. 2007m3).
The different commands I've been using:
. xtreg Return RMRF SMB HML LIQ, re vce(robust)
statsby _b, by(companynum): reg Return RMRF SMB HML LIQ, robust noconstant
The means of my coefficients from statsby are larger than the coefficients from xtreg. Shouldn't the mean of the coefficients from statsby be equal to the coefficients in the other reg commands, or is it wrong to think of it that way?
Please let me know if you find anything that is incorrectly posted and I will correct accordingly.
Many thanks!
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input float(date companynum) double(RMRF SMB HML LIQ) 677 1 -.03567645 .031857625 .035097148 -.0038981533515944945 678 1 .05532546 .0022137638 .0075254366 .0028282892888514353 679 1 .024670409 .0083233304 .0063983239 .048738782080504486 680 1 .017835448 .025180005 -.006078342 .02920958419799065 681 1 -.009465334 .0084097693 .012648277 -.011058678221505731 682 1 .014410426 .0064111673 -.012598168 -.01494227358556188 683 1 .029204069 -.026093747 .018178428 -.014232915409939827 566 2 .052961133 .021395301 -.0039642137 -.036748350946197265 567 2 .064811036 -.02813188 -.015092619 -.017005912917779073 568 2 .016662389 -.046779487 .004310017 .003553448955587013 569 2 -.024995016 .011651467 -.011661232 -.023518709608667893 570 2 -.017945765 .0046412922 .03236885 .0056249091016424176 571 2 -.029406879 -.0016394301 -.0072570862 .025491101902319215 end format %tm date
Comment