Interacting variables

[Joyce Parreira]

Hello, I need some help.

I'm doing my master thesis about renewable energies and politics I have a panel data with 18 years and 155 countries.


I have my regression

xtreg reconsumption lngdppcpppcte2011lag1 gdpgrowthlag1 manufacturinglag1 lnoilpricelag1 energyimportslag1 left presidential absenceofcorruptionlag1, fe vce(robust)

and I wanna know how I can interact

lngdppcpppcte2011lag1

with it owns median. Both variables are continuous.

Thanks

[Carlo Lazzaro]

Joyce:
provided that I cannot follow what you're after, you may want to try:

egen median_var=median(lngdppcpppcte2011lag1)
xtreg reconsumption c.lngdppcpppcte2011lag1##c.median_var gdpgrowthlag1 manufacturinglag1 lnoilpricelag1 energyimportslag1 left presidential absenceofcorruptionlag1, fe vce(robust)

In all likelihood
, you will experience omissions due to collineraity.

[Phil Bromiley]

I'm not sure that this does much - multiplying a variable by a constant just changes the scale but doesn't really change the results. Including a second constant (median_var) simply changes the intercept. Do you mean that you want separate parameters for lngdppcpppcte2011lag1 above and below the median? If so, you would use Carlo's egen then
g highlng=lngdppcpppcte2011lag1>median_var xtreg y highlng##lngdppcpppcte2011lag1 By the way, such long variable names are problematic - often they're substantially truncated in results making it hard to know what variable is what.