regression with no intercept?

[Sandra Bloem]

Hi there,

I am reading a paper where I came across something I didn't understand. I have to do the same research and methodology in Stata as the paper so I was wondering if someone could explain this to me.

They are running this regression (an event study methodology). As you can see it only has dummy variables. Now, I don't see an intercept in this, but I do see an error term (assuming the v is the error term). How does this work? Is is normal to not have a constant if all independent variables are dummy variables or is it not? And do I need to change something in my command?

Thank you in advance

[Chris Boudreaux]

Sandra,

I’m happy to help, but this is not a Stata related question. There are other forums for statistics questions. (https://stats.stackexchange.com/).

The main difference is that including or omitting an intercept changes how you interpret the dummies.

1) Including an intercept: one dummy category will be omitted and each dummy’s effect is interpreted relative to the baseline category (I.e. omitted).
2) Omitting an intercept: each dummy reflects the effect for that category alone.

Most people find 1) more intuitive. Also, are you sure they actually did omit the intercept? They might have carelessly omitted it in the equation only. Most likely the intercept will be in your model by default, but you can drop it if you desire by specifying nocons in the options part of your command.

[Maarten Buis]

The question may not be about Stata, but discussions on statistics are welcom on this forum (and there happens to be a Stata tip on this topic:https://www.stata-journal.com/articl...article=st0250 )

[Sandra Bloem]

Thank you both!

I was puzzling with this since they didn't have nocons after their regression, but they didn't specify this in their framework. I think they omitted it to simplify the look in the text, but in the actual analysis they did put it in.

[Chris Boudreaux]

You're welcome Sandra. Glad you find it helpful.

Maarten, duly noted.