Logistic Regression

Binary outcomes (yes/no) and odds ratios

Back

Model

logit(p) = ln(p/(1-p)) = β₀ + β₁x p = 1 / (1 + e^{-(β₀ + β₁x)})

Exponentiating β gives an odds ratio. If β₁=0.7, OR=exp(0.7)=2.01 (odds roughly doubled per unit x).

Interactive: shape of the logistic curve

-6-3.02
00.802.5
OR per +1 unit x: -
OR = exp(β₁)

Probability vs predictor

Example x could be plaque score, age, or risk index.

Real Dental Scenario: Implant Failure Risk from Smoking

A study models implant failure probability based on the number of cigarettes smoked per day. The fitted logistic model is:

logit(p) = -3.5 + 0.15 × cigarettes/day     OR per cigarette = exp(0.15) = 1.16

Check Patient Risk

02040
0% failure probability 0% 100%
Adjust slider and see risk update in real time
Odds Ratio: Each additional cigarette/day multiplies the odds of failure by 1.16. A 20-cigarette/day smoker has exp(0.15×20)=20× the baseline odds.

S-Curve: Failure Probability vs Cigarettes/Day

The red dot shows the current patient's position on the curve.

Dental example

Outcome: implant failure (yes/no). Predictors: smoking, diabetes, bone quality, clinician experience. Logistic regression estimates adjusted odds ratios for each predictor.