Let’s break this down clearly to resolve your confusion:

1. What do Intercept and Reference (Ref) mean in Logistic Regression? #

Intercept (intercept) #

In binomial logistic regression, the intercept represents the log odds of your target outcome occurring when all predictor variables are set to their reference categories.

For example, if you’re modeling "customer purchase likelihood" with a predictor "location" (Downtown vs. Suburb), the intercept is the log odds of a purchase for customers in the reference location (say, Downtown) when no other predictors are in play. If you have multiple predictors, it’s the baseline log odds when every single predictor is at its reference state.

Reference Group (Ref) #

Categorical predictors (like location) can’t be directly plugged into a regression model—we use dummy coding to convert them into numerical values. One category gets picked as the reference group: this is the baseline we compare all other categories against.

The coefficient for any non-reference category tells you the difference in log odds between that category and the reference group. So if "Downtown" is the reference, the coefficient for "Suburb" would show how much higher or lower the log odds of the outcome are for suburban customers compared to downtown ones.

2. Why does the first row of my model results show intercept with no corresponding predictor? And is Downtown the reference group? #

The intercept row is totally logical #

The intercept isn’t linked to a specific predictor—it’s the baseline log odds when all predictors are in their reference states. That’s why you don’t see a predictor name paired with it in the results. Think of it as the "starting point" of the model before we account for any effects from your predictors.

For example, if your model output looks like this:

The intercept (0.7) is the log odds of the outcome for the reference group (Downtown). The coefficient for Suburb (-0.25) means suburban customers have log odds that are 0.25 lower than downtown customers.

Is Downtown the reference group? #

If your results only show a coefficient for the other location (and not Downtown), then yes—Downtown is definitely the reference group. Regression models only report coefficients for non-reference categories because their values are interpreted relative to the baseline reference. If Downtown weren’t the reference, you’d see its coefficient listed instead (and the other location would be the unreported baseline).

内容的提问来源于stack exchange，提问作者BitwiseStarter