Why does adding variables decrease my degrees of freedom in a regression?

[Rachel Jones]

As I increase the variables in my regression I see my number of observations and degrees of freedom decrease! I'm very perplexed by this as I thought adding variables would have to add degrees of freedom so something is clearly wrong.

I have checked for missing values, but as I learned in my last post, they shouldn't be a concern here as there is a period in front of them so they should be ignored.

Is there another strategy I can use?

Thank you!

[William Lisowski]

Stata ignores missing values, yes. In general, when fitting models it ignores them by ignoring every observation for which one or more variables has a missing value.

So if you add a new variable has missing values for observations that don't have missing values on any of the previous variables, then those observations will now be ignored and the number of observations and the number of degrees of freedom in the regression will decrease.

[Rachel Jones]

Stata ignores missing values, yes. In general, when fitting models it ignores them by ignoring every observation for which one or more variables has a missing value.

So if you add a new variable has missing values for observations that don't have missing values on any of the previous variables, then those observations will now be ignored and the number of observations and the number of degrees of freedom in the regression will decrease.

That makes a lot of sense, thank you!

So I should just delete all my missing values and it will be fine?

[Nick Cox]

No; deleting all your missing values won't help at all. As Stata those ignores missing values, you won't get different results by deleting the corresponding observations. The only way to change your situation is to impute the missing values with non-missing values, much easier said than done.