number of cases for regression

#1
Hi guys,

Is there a minimum number of cases in linear regression? I did regression for 20 countries in 5 different fiscal years so together I had 100 cases. However, variables that I had expected to be significant appeared to be insignificant. Do you think results may significantly change if I add more cases, eg. 100 more?

thanks,
Zofia
 
#2
I think the answer to this question depends on who you ask. Some will say that 10 values per predictor is enough, others will say that 15, 20 or even 30 are needed. It will also depend on other factors such as anticipated effect sizes...but as a rule of thumb, the more predictors per variable; the better.

Not totally sure what your IV/DV in your model are, but if I understand correctly and you only have 5 values per country, it may be partially responsible for lack of significance.
Hope this helps.
 
#3
Hi Zofia,
yeah, it is true. it depends on whom you ask. Though, a thumb rule is you should at least have 20 variables (in Economics/Finance) or 30 variables (Social Science/Marketing)

The reason of unpredicted result may also be due to Panel regression, rather than the sample size.
since you have 5 different country's data, it is possible that the hypothesis you are arguing is valid for some and invalid or non-supportive for the others.