The Logit Link Function and Odds
What you'll learn: How the logit function transforms probabilities to the log-odds scale, enabling linear modeling of binary outcomes.
Why Transform Probabilities?
In logistic regression, we model binary outcomes (yes/no, 0/1). Probabilities range from 0 to 1, but linear models need to work on an unlimited scale (-∞ to +∞). The logit link function solves this problem by transforming probabilities into something that can range freely: log-odds.
From Probability to Odds
First, let's understand odds. If the probability of an event is p, the odds are:
Odds = p / (1 - p)
For example, if p = 0.75 (75% chance), the odds are 0.75/0.25 = 3. We say "3 to 1 odds" — three times more likely to happen than not happen.
Odds range from 0 to ∞ (but still not -∞ to +∞).
The Logit Transformation
The logit function takes one more step — it applies the natural logarithm to the odds:
logit(p) = log(p / (1 - p))
This log-odds now ranges from -∞ to +∞, perfect for linear modeling! When p = 0.5, logit = 0. When p > 0.5, logit is positive. When p < 0.5, logit is negative.
Why It Matters
The logit link allows us to write:
log(p / (1 - p)) = β₀ + β₁X₁ + β₂X₂ + ...
Now we have a linear relationship between predictors and the log-odds of the outcome, which we can estimate using standard techniques you've learned.
Key Takeaway: The logit link function transforms probabilities into log-odds, creating an unbounded scale that allows linear modeling of binary outcomes through the odds ratio framework.