Understanding/Reading Regression Output With Categorical Variables

[Paige LaPierre]

Hello;

I have experience with regression, I just haven't done it in a while and am SUPER confused at the moment, and thought it would be worth it to ask; its a really easy question.



I am doing a regression analysis, and some of my control variables are categorical values with values ranging from 1-4.

For example, one variable is health, where 1=excellent, 2=good, 3=fair, and 4=poor

If I were to run a regression, what value would the coefficient use as a reference?

Would I be reading it as if health is excellent, or poor? Etc?

Thanks!

[George Ford]

You can't use a single variable. You need dummies for 3 of the 4, thus excluding 1 to avoid the dummy trap. I'd exclude "poor," so all dummies are in reference to "poor".

could use i.health as a variable, or b4.health which sets "poor" as the base.

[William Lisowski]

You should definitely follow George's suggestion to use Stata's factor variable notation rather than create separate indicator (dummy) variables. If you are unfamiliar with factor variable notation, consult the output of

help fvvarlist

[Paige LaPierre]

Thank you!

[Weiwen Ng]

You can't use a single variable. You need dummies for 3 of the 4, thus excluding 1 to avoid the dummy trap. I'd exclude "poor," so all dummies are in reference to "poor".

could use i.health as a variable, or b4.health which sets "poor" as the base.

Speaking of that, Stata will always use the lowest number as the base category. You can manually tell it to use a certain category, e.g. as already stated, ib4.health will tell Stata to use level 4 as the base.

say you decide that your coding scheme isn’t what you really wanted. The recode command is a nice shortcut.