[Linear Regression] Regression with aggregated or separated years?

[andre94]

Hello to everyone! :wave:

I should do some regressions but i've an important and simple question...

my dataset have this structure:

|geographic_area|year|sales|popolation|
Spain 2000 40 1000
Spain 2001 30 2000
Spain 2002 20 3000
Italy 2000 40 1000
Italy 2001 30 2000
Italy 2002 20 3000

I should investigate the relation between sales and popolation but i don't know if regress by year or not. If i do (sales regress to popolation) i consider the areas and the years aggregated but this make sense? or it is preferible to consider years separately?

Thanks a lot :tup:

 

[hlsmith]

Is this your full dataset or are there more years and countries?

 

[andre94]

Hello! Thanks for your reply!!

No i've have much more countries and much more years... that is only an example for understanding the logic behind

 

[hlsmith]

Hmmm, you should know your data best, but I would not see why you wouldn't control for year. If there is volatility between years you would want to collect that info. Also, if effects are not constant across year and country, etc., you may need to examine interaction terms.

 

[andre94]

Thanks for your reply

yeah, as you saying, maybe the effects of the years will lost. I would do that for enlarge a dataset that have few observation by years (10 countries) and i demand to myself if I catch the same effects but with more elegancy in the graphs that I produce (for example).

I'm a little bit confused but i will operate with years separately despite of the ten observations by years.

Instead, how implement interactions terms? T

Thanks a lot,
Andrea