This is a lagging analysis, following students that began in 2012-2013, 2013-2014, 2014-2015. A cohort includes first time freshman and transfer students that begin in a given year, with students starting in the summer counted toward the next fall. Only students that declare a major are counted. Graduation is considered up to current time.
All three factors have main effects (via Logistic Regression) and an interaction between race/ethnicity and generation. In other words, being first generation for a represented student raises the probability of not graduating a lot. However, if one is an underrepresented student, the deck is already stacked against you and being first generation does not increase the probability of not graduating as much.
Gender has a uniform effect regardless of generation and race/ethnicity.
## Analysis of Deviance Table (Type II tests)
##
## Response: cbind(n, total)
## LR Chisq Df Pr(>Chisq)
## gender 65.522 1 5.746e-16 ***
## race_ethnicity 45.845 1 1.280e-11 ***
## generation 93.282 1 < 2.2e-16 ***
## gender:race_ethnicity 0.858 1 0.35435
## gender:generation 2.938 1 0.08649 .
## race_ethnicity:generation 4.040 1 0.04444 *
## gender:race_ethnicity:generation 0.027 1 0.86923
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1