How many observations are sufficient for an OLS regression?

[Bright Tree]

Dear friends,

How many observations are sufficient for an OLS regression? If I only have around 70 obs, was the estimation result reliable?

[Adrien Bouguen]

I don't think this is a stata question rather an econometrics question...

Above 36 observation, the normality assumption is said to be met. So in principle, 70 is sufficient to insure some of the basics assumption of the OLS. Yet the reliability of the OLS depends only weakly on the sample size. Other assumption such as the 0CM/OVB assumption are more crucial...

[Bright Tree]

Dear Professor Bouguen, Thank you a lot for your help!

[Carlo Lazzaro]

Bright:
as an aside to previous helpful comments, we should also keep in mind that Portnoy's suggested ratio (https://projecteuclid.org/euclid.aos/1176346793) states that [(constant+predictors)^2]/sample size should ideally go to 0 as the sample size goes to infinity.
Eventually, in OLS normality is a (weak) requirement for the residual distribution only.

[Bright Tree]

Dear Professor Lazzaro, thank you for your help!

[Carlo Lazzaro]

Bright:
Carlo is enough. Thanks.

[Marcos Almeida]

With small sample size, the number of predictors will be limited. That said, maybe a different approach - help bayes - will be helpful.