Hello statalist,
I am researching the effect of chief digital officers on strategic change at a company by doing a GEE regression.
One of the hypotheses that I test is: The association between CDO presence and strategic change is stronger if the CDO is part of a company’s executive committee.
My panel dataset looks as follows. The dummy variable for executive committee is equal to one if the executive is part of the committee, zero if not, and empty if there is no chief digital officer in that firm-year.
When I implement it as below it ommits the direct effect of the variable cdo and of the interaction term between cdo and execCommittee. It measures only the direct effect of the execCommittee dummy, which makes sense as the variable is empty for all firm-years without cdo and therefore the regression runs only with the observations that have no missing values.
I am unsure whether this is the right implementation to test the above hypothesis. When I implement it differently (putting a zero instead of a missing value in any case, except from execComittee = 1, meaning no missing values in that variable), then the interaction term is ommitted because of collinearity.
I think the issue here is that there is no observation where cdo = 0 and execCommittee = 1. But I don't know how to test the hypothesis then.
Thanks in advance for your help and advice.
I am researching the effect of chief digital officers on strategic change at a company by doing a GEE regression.
One of the hypotheses that I test is: The association between CDO presence and strategic change is stronger if the CDO is part of a company’s executive committee.
My panel dataset looks as follows. The dummy variable for executive committee is equal to one if the executive is part of the committee, zero if not, and empty if there is no chief digital officer in that firm-year.
Code:
* Example generated by -dataex-. For more info, type help dataex clear input long GVKEY double year float(cdo dummy_execComittee stratchange1) 1045 2010 0 . -2.601784 1045 2011 0 . -2.0886009 1045 2012 0 . -2.873147 1045 2013 0 . -1.2799462 1045 2014 0 . -.3453558 1045 2015 0 . -1.2935414 1045 2016 0 . -1.763744 1045 2017 0 . -3.0115 1045 2018 0 . 1.2935827 1045 2019 0 . 1.2720878 1075 2010 0 . -2.7677424 1075 2011 0 . -2.865253 1075 2012 0 . -2.969077 1075 2013 0 . -2.9866655 1075 2014 0 . -2.7382064 1075 2015 0 . -2.9036155 1075 2016 0 . -2.7266235 1075 2017 0 . -2.862787 1075 2018 0 . -2.978583 1075 2019 0 . -2.623896 1078 2010 0 . -2.2792287 1078 2011 0 . -2.642496 1078 2012 0 . -2.967848 1078 2013 0 . -1.9837472 1078 2014 0 . -2.9220934 1078 2015 0 . -2.764499 1078 2016 0 . -2.390353 1078 2017 0 . -2.5110555 1078 2018 0 . -2.864998 1078 2019 0 . -2.858804 1161 2010 0 . -.9779071 1161 2011 0 . -2.132206 1161 2012 0 . -1.481019 1161 2013 0 . -2.2021053 1161 2014 0 . -.7882497 1161 2015 0 . -.10412782 1161 2016 0 . -.8612247 1161 2017 0 . -1.9829502 1161 2018 0 . -2.1478033 1161 2019 0 . -2.3022478 1209 2010 0 . -2.9827795 1209 2011 0 . -3.0681744 1209 2012 0 . -2.670916 1209 2013 0 . -2.6834695 1209 2014 0 . -2.852429 1209 2015 0 . -2.866161 1209 2016 0 . -2.8276665 1209 2017 0 . -2.5355625 1209 2018 0 . -3.042375 1209 2019 0 . -2.790608 1230 2010 0 . -2.5241604 1230 2011 0 . -2.799176 1230 2012 0 . -2.786882 1230 2013 0 . -2.795616 1230 2014 0 . -2.962554 1230 2015 0 . -2.9254184 1230 2016 0 . -2.481045 1230 2017 0 . -2.7610176 1230 2018 0 . -2.9477084 1230 2019 0 . -2.851113 1300 2010 0 . -2.806876 1300 2011 0 . -2.764426 1300 2012 0 . -2.892504 1300 2013 0 . -2.981084 1300 2014 0 . -2.958081 1300 2015 0 . -2.9666114 1300 2016 0 . -2.558535 1300 2017 0 . -3.11228 1300 2018 0 . -2.816524 1300 2019 0 . -2.7739556 1327 2010 0 . -2.691804 1327 2011 0 . -2.95882 1327 2012 0 . -2.5993896 1327 2013 0 . -2.898453 1327 2014 0 . -2.931638 1327 2015 0 . -2.480188 1327 2016 0 . -2.428869 1327 2017 0 . -2.938254 1327 2018 0 . -2.670571 1327 2019 0 . -2.0389254 1380 2010 0 . -2.8321004 1380 2011 0 . -2.858678 1380 2012 0 . -2.8684785 1380 2013 0 . -2.714155 1380 2014 0 . -2.525053 1380 2015 0 . -2.486863 1380 2016 0 . -2.705614 1380 2017 0 . -2.571207 1380 2018 0 . -2.983559 1380 2019 0 . -2.853369 1440 2018 1 0 -2.887975 1440 2019 1 0 -2.719169 1440 2010 0 . -2.624753 1440 2011 0 . -2.717351 1440 2012 0 . -2.800929 1440 2013 0 . -2.824792 1440 2014 0 . -2.67811 1440 2015 0 . -3.0011826 1440 2016 0 . -2.528976 1440 2017 0 . -3.033558 end
When I implement it as below it ommits the direct effect of the variable cdo and of the interaction term between cdo and execCommittee. It measures only the direct effect of the execCommittee dummy, which makes sense as the variable is empty for all firm-years without cdo and therefore the regression runs only with the observations that have no missing values.
Code:
xtgee stratchange1 L.cdo##L.dummy_execComittee, family(Gaussian) link(identity) corr(ar) vce(robust)
I think the issue here is that there is no observation where cdo = 0 and execCommittee = 1. But I don't know how to test the hypothesis then.
Thanks in advance for your help and advice.
