OLS regresion at individual and group level. Collapse?

[Diego PF]

Good morning one and all.

I have a database with information for 40 urban regions. Within each region, there are a different number of cities. I want to do an OLS regression at the scale of urban regions (not cities), so I have to "collapse" (I don't know if that's the correct term) the data from the database because each row in the database (each city ) repeats the data from the urban region.

What I have done is use:
collapse (mena) var_list , by (id_urbanregion)

And then run the regression with OLS:
reg var1_region var2_region region_var3 region_var4 region_var5 , r

Is it correct, or is there another way to do it? Clearly, results between individuals (cities) and groups (urban regions) are completely different, so I am unsure if my procedure is OK using STATA commands.

Could you give me some advice? I really appreciate it!

Best,
Diego

[Carlo Lazzaro]

Diego:
your code (-mena- reads -mean-) is correct.
Obviously the two regressions give back different results, as the latter investigates data aggregated at regional level.

[Diego PF]

Carlo, thanks for your kind answer.
Considering that each urban region has a different number of cities, do I need to wight somehow when I collapse my data? Sorry for bothering you with these basic questions, but I did not find a clear answer in the books.
Thanks,
Diego

[Carlo Lazzaro]

Diego:
yes,
See -[fw=number]- examples available from -collapse- helpfile.