model usage for prediction

#1
Hi everone,
I am facing a twofold problem. First, can i cross sectional model be used for prediction? Mine is a sales regression for 2011 and I want to predict the changes in the 2012, so its a log log model. Second, I choose a buch of explanatory variables like gdp p.c, population, HDI, IEF etc. to capture changes in buying power usw. Is there a stronger indicator that can explain changes in the total potential sales (not market share!)?
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
Cross -sectional design have limitations establishing causality, you probably can, more for hypothesis generation then testing. If rationale is strong and well established you will have some ground to stand on, don't know of your scenario content to speak beyond this.
 
#3
Cross -sectional design have limitations establishing causality, you probably can, more for hypothesis generation then testing.
From what I understand causality tests are indeed applied in time series e.g Granger causality, but your suggestion is to simply proceed to hypothesis tests regarding the generated coefficients e.g F-test right?
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
I was saying that you can do it, just keep in mind the limitations when making any conclusions and these data should guide future more rigorous examinations on this topic.