Weighted Sum of Regression Coefficients

[Nicholas Hamer]

Hi,

I have conducted 6 different regressions. The resulting coefficients for the variables Pre4, Pre3, Pre2, Pre1, t , Post1, Post2, Post3, and Post4 have been stored in a table using esttab. I am wondering how I can sum these coefficients across the 6 regressions , weighted by the inverse of their standard errors to obtain pooled coefficient estimates. I have searched online and cannot see a simple way to do this. Each of the regressions is the result for a different host country, with my overall research looking at the effects on the growth rate for a country hosting the football world cup.

Any help is much appreciated. A copy of the table is attached.

Thanks.

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

You can save the parameters and then do the calculations in Stata. While some write the parameters to a new data set, I often just create a variable filled with missing values, and the replace the contents in a specific observation number with the parameter. AI don't know esttab, but I'd be surprised if it includes the arithmetic you want.

[Nicholas Hamer]

Thanks, sorry this is my first time posting so I wasn't entirely sure what to include.

I have run 6 regressions on 6 different samples. Each sample consists of an 8 year window around a World Cup year, including the GDP and calculated GDP growth rates for the host country for that particular World Cup and the other countries that bid to host.

This is the example code for one of the 6 regressions.

*South Korea/Japan (2002)
clear
use "CountryPanelData  1990-2016"
keep if inlist(Country, 128,105,99)
drop if year < 1997
drop if year >2006
reg OwnGDPGROWTH Pre4 Pre3 Pre2 Pre1 t Post1 Post2 Post3 Post4 i.year i.Country , cluster(CountryName)
eststo SKJap

After doing this for each sample, I call the regression estimates using:

esttab, keep(Pre4 Pre3 Pre2 Pre1 t Post1 Post2 Post3 Post4)

The regression estimates are then shown as in the original post (but with 6 columns instead of 5).

Example data is shown below.

* Example generated by -dataex-. To install: ssc install dataex
clear
input str30 CountryName int year str3 CountryCode double(GDPCAP GDP) long(Population Country) float(Host Pre4 Pre3 Pre2 Pre1 t Post1 Post2 Post3 Post4 Balance OwnGDPGROWTH)
"Japan"       1997 "JPN" 41861.753438585794   5276967053207.81 87033835  99 1 0 0 0 0 0 0 0 0 0 1  1.0702133
"Japan"       1998 "JPN"  41277.06675880894   5217421238313.45 86946840  99 1 1 0 0 0 0 0 0 0 0 0 -1.1348724
"Japan"       1999 "JPN"  41097.95960397278  5204275722610.677 86762009  99 1 0 1 0 0 0 0 0 0 0 0  -.2521515
"Japan"       2000 "JPN"  42169.73328377276  5348935478913.588 86547755  99 1 0 0 1 0 0 0 0 0 0 0   2.741623
"Japan"       2001 "JPN"  42239.18492637218  5370670124203.297 86351962  99 1 0 0 0 1 0 0 0 0 0 0   .4055023
"Japan"       2002 "JPN" 42190.804872800196 5377007127014.0205 86157117  99 1 0 0 0 0 1 0 0 0 0 0  .11787415
"Japan"       2003 "JPN"  42744.01128499001 5459179633296.3545 85931170  99 1 0 0 0 0 0 1 0 0 0 0  1.5167236
"Japan"       2004 "JPN" 43671.679973901904  5579537505145.682 85503611  99 1 0 0 0 0 0 0 1 0 0 0  2.1806717
"Japan"       2005 "JPN" 44393.626384213574  5672306823990.121 84991644  99 1 0 0 0 0 0 0 0 1 0 0  1.6490936
"Japan"       2006 "JPN" 44995.494491919875  5752853952769.924 84556191  99 1 0 0 0 0 0 0 0 0 1 0   1.409912
"Korea, Rep." 1997 "KOR" 13481.626171284572  619528986792.2192 32782525 105 1 0 0 0 0 0 0 0 0 0 1   5.753517
"Korea, Rep." 1998 "KOR" 12652.353485338568  585633197556.1841 33210599 105 1 1 0 0 0 0 0 0 0 0 0  -5.626678
"Korea, Rep." 1999 "KOR" 13983.412829036064  651860239208.8304 33594533 105 1 0 1 0 0 0 0 0 0 0 0  10.713768
"Korea, Rep." 2000 "KOR"  15104.52151764616  710035024103.3992 33939664 105 1 0 0 1 0 0 0 0 0 0 0   8.548355
"Korea, Rep." 2001 "KOR" 15667.378459426205  742166287073.0867 34285089 105 1 0 0 0 1 0 0 0 0 0 0  4.4260025
"Korea, Rep." 2002 "KOR" 16734.845664021348  797327303663.0418 34482015 105 1 0 0 0 0 1 0 0 0 0 0   7.169151
"Korea, Rep." 2003 "KOR"  17136.66156711312  820714650870.4987 34615235 105 1 0 0 0 0 0 1 0 0 0 0   2.890968
"Korea, Rep." 2004 "KOR" 17905.225791449895  860928359316.6796 34731974 105 1 0 0 0 0 0 0 1 0 0 0    4.78363
"Korea, Rep." 2005 "KOR" 18568.362814706383  894708410715.3514 34841188 105 1 0 0 0 0 0 0 0 1 0 0   3.848648
"Korea, Rep." 2006 "KOR" 19427.189841436506  941019894278.9352 35090058 105 1 0 0 0 0 0 0 0 0 1 0   5.046654
"Mexico"      1997 "MEX"   8296.16586148783  807065442037.9694 58205025 128 0 0 0 0 0 0 0 0 0 0 1   6.622505
"Mexico"      1998 "MEX"  8588.637851290176  848741697521.2067 59407888 128 0 0 0 0 0 0 0 0 0 0 1   5.035019
"Mexico"      1999 "MEX"  8694.987299958215  872112260583.4557 60582113 128 0 0 0 0 0 0 0 0 0 0 1   2.716446
"Mexico"      2000 "MEX"  8997.433617416287  915216005402.7919 61745507 128 0 0 0 0 0 0 0 0 0 0 1   4.824066
"Mexico"      2001 "MEX"  8843.901159961495  911514962239.0303 62890102 128 0 0 0 0 0 0 0 0 0 0 1 -.40512085
"Mexico"      2002 "MEX"  8731.220020574438   911151773828.818 64023490 128 0 0 0 0 0 0 0 0 0 0 1 -.03986359
"Mexico"      2003 "MEX"  8749.778035387033  924330515307.7362 65169756 128 0 0 0 0 0 0 0 0 0 0 1  1.4360428
"Mexico"      2004 "MEX"  8977.657821141815  960569732547.5781 66355990 128 0 0 0 0 0 0 0 0 0 0 1   3.845596
"Mexico"      2005 "MEX"  9059.810486315213  982737828708.3746 67605637 128 0 0 0 0 0 0 0 0 0 0 1  2.2815704
"Mexico"      2006 "MEX"  9327.736199909536 1026912659604.7242 69062443 128 0 0 0 0 0 0 0 0 0 0 1   4.397011
end

I am looking for how to weight the coefficients by the the inverse of their standard errors, to produce a pooled estimate for each dummy variable (Pre 4 ..... Post4). I also want to obtain the standard errors for these pooled estimates.

I am hoping this is now clearer.