Optimal number of observations in OLS

[j235]

How many observations is optimal to use for OLS ? In reality I often face with time series which contains 6-7 observation with yearly data.So my question is does this regression is relevant with this small number of observation?

[startz]

How many observations is optimal to use for OLS ? In reality I often face with time series which contains 6-7 observation with yearly data.So my question is does this regression is relevant with this small number of observation?

Probably not, although the answer depends on circumstances of course.

[jmagomez]

Dear j235,

As told by startz, it depends on the circumstances, but from my experience I believe that if your database is a monthly basis, it would take around 120 observations, like industrial production. If they are annual, at least 25 to 30 observations, like gdp.

Regards.

[j235]

Thank you.
This is very useful for me.
But what happened with quarterly data ?
How many observations is enough for relevant regression ?

[startz]

There really aren't any rules. jmagomez has made some reasonable observations, but even these may or may not be right depending on what you're studying and on how many variables are included.

[jmagomez]

Dear j235,

Startz is totally correct. There are no rules. The number of observations that I told you follows my experience.

For Quarterly data, I think that at least 40 observations.

Regards.