Dummy Variable

[Teiren Mistry]

Hello! Appreciate if anyone has any advice.

I am doing a logit model. I have generated dummy variables for some of my variables like ethnicity, splitting it into white, black/mixed and asian. I wanted white to be my reference, so do I not include this in my regression? When I tried this using i. it put a 0. next to mixed/black saying this needs to be omitted. I tried this both with and without including the white variable as I know there's the n-1 assumption. However both times it said mixed/black needs to be omitted. Is there a way I can choose my reference category, as I wanted to compare other ethnicities to that of white individuals.

[Dirk Enzmann]

Welcome to the Stata Form.

Please read the complete Forum FAQ before posting (especially #12) and provide us with an example of your data. Also: Don't post a screenshot of your results table but use code delimiters, instead.

[Ken Chui]

Logistic regression cannot deal with any subgroup that results in perfect prediction (aka, outcomes is all 1, or outcomes are all 0.) This is called "complete separation". You can learn more about this phenomenon here: https://stats.oarc.ucla.edu/other/mu...ith-the-issue/