Replace missing values with values from other observations

[Steffen Weisthoff]

Hello,

I´m stucked with my data cleaning. Below, a excerpt of my dataset is shown. I only have included data from one gvkey code, but in total there are approx. 7500 unique gvkey codes in my dataset.

As you can see below, I only have some values for my co_per_rol and becameceo variable. In case of missing values, I want to have the same values like for the observations with nonmissing values.

I tried to solve this problem by using total command, however this does not work if I have more than 1 nonmissing values for the observations.

Thanks in advance!

* Example generated by -dataex-. For more info, type help dataex
clear
input long gvkey int fyearqtr_ec long co_per_rol int becameceo
1045 217     .     .
1045 223     .     .
1045 220     .     .
1045 221     .     .
1045 222     .     .
1045 227     .     .
1045 224 49256 19701
1045 225     .     .
1045 226     .     .
1045 231     .     .
1045 228     .     .
1045 229     .     .
1045 230     .     .
1045 235     .     .
1045 232 49256 19701
1045 233     .     .
1045 234     .     .
1045 236     .     .
1045 237     .     .
1045 238     .     .
1045 243     .     .
1045 240     .     .
1045 241     .     .
1045 242     .     .
1045 247     .     .
1045 244     .     .
1045 245     .     .
1045 246     .     .
end
format %tq fyearqtr_ec
format %td becameceo

[Joro Kolev]

You might want to think more carefully what you want to do here. For the example you are showing, the person became CEO before the data span started, so what you are asking makes sense for this co_per_rol... Whether it would make sense in general has a question mark on it.

For the sample that you are showing,

. egen whenceo = max(becameceo), by(gvkey)

. format whenceo %td

would do the job. But this is NOT generally true if you have repeated CEOs by company.