Predict outcomes
Estimate pocket depth from age, smoking status, and plaque levels.
Dental research
Identify which risk factors truly influence implant failure or caries.
Report results
Present coefficients, odds ratios, and confidence intervals in papers.
Use when the outcome is continuous, like attachment loss, plaque score, or healing time, and you want to estimate how much it changes with a predictor.
Use when several predictors may influence the same continuous outcome and you want to adjust for confounding variables at the same time.
Choose this when the outcome is binary, such as implant failure yes/no or peri-implantitis present/absent.
A model is not judged only by p-values. You also need to know how well it explains variation and whether the assumptions are believable.
Click a card to learn and try it.
Click the chart to add data points, then fit a regression line. Try preset dental datasets.
Adjust the slider to see how smoking affects implant failure probability.
Watch a regression study unfold step-by-step: recruit patients, collect data, fit the model, interpret results.
Linear Regression
Report: B (95% CI), p-value
e.g. B=0.04 (0.02-0.06), p=0.001
Logistic Regression
Report: OR (95% CI), p-value
e.g. OR=2.5 (1.3-4.8), p=0.006
Model Summary
Report: R², Adjusted R², F-stat
e.g. R²=0.42, F(3,96)=23.1, p<0.001
| Model Type | Coefficient | 95% CI | p-value | Interpretation |
|---|---|---|---|---|
| Linear Age → CAL | B = 0.04 | 0.02 – 0.06 | 0.001 | Each year of age increases CAL by 0.04 mm |
| Multiple Age + Smoking → PD | Bage=0.03 Bsmoke=1.2 | 0.01–0.05 0.6–1.8 | 0.004 <0.001 | Adjusted for confounders |
| Logistic Smoking → Implant Failure | OR = 2.5 | 1.3 – 4.8 | 0.006 | Smokers have 2.5x odds of implant failure |
"Multiple linear regression was performed to examine the effect of age, smoking status, and plaque index on clinical attachment loss. The model was statistically significant, F(3, 96) = 23.1, p < 0.001, R² = 0.42. Age (B = 0.04, 95% CI: 0.02–0.06, p = 0.001) and smoking (B = 1.2, 95% CI: 0.6–1.8, p < 0.001) were significant predictors, while plaque index was not (B = 0.15, 95% CI: -0.1–0.4, p = 0.22)."
Continuous outcome. Y = B0 + B1*X. Report B with 95% CI and p-value. Check linearity, normality of residuals.
Multiple predictors. Adjusts for confounders. Check multicollinearity (VIF < 10). Report adjusted R².
Binary outcome (yes/no). Reports odds ratios. Use when predicting implant failure, disease presence.
OR = 1 (no effect), OR > 1 (increased risk), OR < 1 (protective). Always report with 95% CI.
Independence, linearity, no multicollinearity, adequate sample size. EPV >= 10 for logistic regression.
R² = proportion of variance explained. Higher is better. Adjusted R² penalizes adding useless predictors.
The sign shows direction and the size shows expected change in the outcome for a one-unit increase in the predictor, holding other variables constant.
In logistic models, OR greater than 1 means higher odds, OR less than 1 means lower odds. It describes multiplicative change, not direct probability change.
The interval shows precision. Wide intervals suggest more uncertainty, even if the p-value looks interesting.
These tell you how much outcome variability is explained by the model, but a high value alone does not prove causation or good clinical usefulness.