Permutation Logistic Regression Calculator

Enter Dataset and Settings

Outcome Column

Use a header name or a 1-based column number.

Predictor Columns

Use comma-separated names or numbers. Up to three predictors.

Prediction Values

Provide one value for each predictor to generate a probability.

Delimiter

Permutation Runs

Maximum Iterations

Convergence Tolerance

First row contains headers

Dataset

Outcome must be binary and coded as 0 or 1. Predictors must be numeric.

Example Data Table

Outcome	StudyHours	PrepScore	Attendance
1	8	82	92
1	7	76	88
0	3	45	60
0	4	52	66
1	9	91	95
0	5	58	70

You can replace this with your own binary outcome dataset. Keep the outcome coded as 0 or 1.

Formula Used

Logistic model: logit(p) = β0 + β1x1 + β2x2 + ... + βkxk, where p = 1 / (1 + e^(-logit(p))).

Odds ratio: OR = e^β. A one-unit increase in a predictor multiplies the odds by its odds ratio, holding other predictors fixed.

Permutation test: The calculator repeatedly shuffles the binary outcome, refits the model, and measures how often shuffled coefficients or model fit are at least as extreme as the observed result.

Fit statistics: AIC = 2k - 2LL, BIC = ln(n)k - 2LL, and McFadden R² = 1 - (LL model / LL null).

How to Use This Calculator

Paste your dataset into the text area using comma, semicolon, or tab separation.
Identify the binary outcome column and up to three numeric predictor columns.
Set the number of permutation runs, iteration limit, and convergence tolerance.
Optionally enter predictor values to generate a predicted probability.
Press Run Calculator to show results above the form.
Review coefficients, odds ratios, confidence intervals, and permutation p-values.
Use the CSV and PDF buttons to export the computed output.

Frequently Asked Questions

1. What does permutation logistic regression measure?

It estimates a binary logistic model and validates effects by shuffling outcome labels many times. This gives empirical p-values that are often more robust than relying only on asymptotic assumptions.

2. Why use permutation p-values?

Permutation p-values compare observed coefficients against shuffled datasets. They help when sample sizes are moderate, distributions are unusual, or classical significance tests may be overly optimistic.

3. How many predictors should I include?

This calculator supports up to three predictors for stability and speed. With small samples, fewer predictors usually give more trustworthy estimates and smoother convergence.

4. What do odds ratios mean here?

An odds ratio shows how the odds of outcome equals one change for a one-unit increase in a predictor, while other included predictors stay fixed.

5. What if my model does not converge?

Try fewer predictors, more rows, a cleaner dataset, or scaled variables. Separation or nearly perfect prediction can make logistic models unstable and inflate coefficients.

6. Can I use categorical predictors?

Only after converting categories into numeric dummy variables before pasting data. Each resulting indicator column can then be used like any other predictor.

7. What does the global permutation p-value show?

It compares observed model log-likelihood against models fit after shuffled outcomes. Smaller values suggest the overall model captures a real signal beyond random label assignments.