Logistic Regression Calculator

It's not a rounding coincidence — the engine's Odds Ratio formula, const oddsRatio = Math.exp(beta1), is written to depend on β₁ alone. Neither β₀ nor the Predictor Value appears anywhere in that line, so changing either one, by any amount, leaves the Odds Ratio completely unchanged; only β₁ ever moves it.

The logit is computed as beta0 + beta1 * xValue, and only the product of β₁ and x feeds into the sigmoid function — the two factors never appear separately. Scaling either one by the same percentage changes that product by the same amount, so a small nudge to either input shifts Predicted Probability by an equal measured amount at any given starting point.

No. It assumes you already have fitted coefficients — β₀ and β₁ — from some other source (statistical software, a published study, a prior analysis) and simply evaluates the resulting sigmoid model at one Predictor Value you supply. There is no data-fitting step anywhere in the engine; it is a pure evaluator, not an estimator.

Yes — both are wrapped in Math.min(value, 99999) before rounding, as a display safety net against runaway exponentials. For Odds Ratio that ceiling never actually engages within the calculator's own allowed β₁ range of -10 to 10, since e^10 is only about 22,026 — well under the 99,999 cap even at the most extreme coefficient the input field permits. Odds is not protected the same way: it depends on the logit (β₀ + β₁ × x), and because β₀ can range to ±20 and the Predictor Value to ±1000, the logit can easily exceed roughly 11.5, at which point Odds does hit the cap and displays the clamped 99,999 rather than the true, larger value.