Hi Everyone,
One more question for the day. If I would like to estimate if there is a significant difference of height on weight between male and female, the usual model would be:
xi: regress weight i.female*height The statistical significance of the coefficient of the interaction term and the dummy for gender will tell me if female height effects weight more or less than male. Here is the complete example (http://www.ats.ucla.edu/stat/stata/faq/compreg2.htm) My question is: How can I do this with panel data, fixed effect? the variable gender will be the same for each id and year (xtset id year) and indeed should be dropped out. Any suggestion? I know that in that case random effect should work, but the Hausman test strongly suggest fixed effect. The only thing that I could think of is running a regression for each subgroup 1) xtreg weight height if gender=1, fe
2) xtreg weight height if gender=0, fe
This won't estimate if there is a statistical difference between groups but it will look at the effect of height for ecah group. Thank you Marco
One more question for the day. If I would like to estimate if there is a significant difference of height on weight between male and female, the usual model would be:
xi: regress weight i.female*height The statistical significance of the coefficient of the interaction term and the dummy for gender will tell me if female height effects weight more or less than male. Here is the complete example (http://www.ats.ucla.edu/stat/stata/faq/compreg2.htm) My question is: How can I do this with panel data, fixed effect? the variable gender will be the same for each id and year (xtset id year) and indeed should be dropped out. Any suggestion? I know that in that case random effect should work, but the Hausman test strongly suggest fixed effect. The only thing that I could think of is running a regression for each subgroup 1) xtreg weight height if gender=1, fe
2) xtreg weight height if gender=0, fe
This won't estimate if there is a statistical difference between groups but it will look at the effect of height for ecah group. Thank you Marco