Calculator Input
Paste a header row followed by numeric rows. Time must be positive. Event must be 0 or 1.
Example Data Table
This sample matches the default input. You can paste it directly into the calculator.
| time | event | age | treatment | biomarker |
|---|---|---|---|---|
| 5 | 1 | 62 | 1 | 2.4 |
| 8 | 1 | 55 | 0 | 1.8 |
| 12 | 0 | 47 | 0 | 1.5 |
| 4 | 1 | 71 | 1 | 3.1 |
| 15 | 0 | 50 | 1 | 1.2 |
| 9 | 1 | 66 | 0 | 2.9 |
| 11 | 1 | 58 | 1 | 2.0 |
| 6 | 1 | 73 | 0 | 3.5 |
| 13 | 0 | 45 | 0 | 1.1 |
| 7 | 1 | 60 | 1 | 2.6 |
| 10 | 0 | 52 | 0 | 1.7 |
| 14 | 1 | 68 | 1 | 3.0 |
| 16 | 0 | 49 | 1 | 1.4 |
| 18 | 1 | 64 | 0 | 2.7 |
Formula Used
Cox model: h(t|x) = h0(t) × exp(β1x1 + β2x2 + ... + βpxp)
The baseline hazard h0(t) stays unspecified, while covariates shift the hazard multiplicatively through exp(βx).
Partial log-likelihood:
ℓ(β) = Σi:event [βxi − log(Σj∈R(ti) exp(βxj))]
Hazard ratio: HR = exp(β)
A one-unit increase in a covariate multiplies the hazard by exp(β), holding the other covariates constant.
How to Use This Calculator
- Paste a CSV style dataset with a header row.
- Set the time and event column names exactly as they appear.
- Use 1 for events and 0 for censored observations.
- Keep all covariate columns numeric for estimation.
- Choose the confidence level, iterations, tolerance, and display precision.
- Optionally enter a custom covariate profile using name=value pairs.
- Run the model and review hazard ratios, intervals, fit statistics, and the survival graph.
- Download the coefficient table and summary using the CSV or PDF buttons.
Frequently Asked Questions
1) What does the event column represent?
Use 1 when the event happened and 0 when the case was censored. Censored rows still contribute follow-up time before leaving the risk set.
2) What does a hazard ratio above one mean?
A hazard ratio above one suggests higher event intensity as that covariate increases. Values below one suggest a protective or slower event pattern.
3) Do I need evenly spaced follow-up times?
No. Cox regression works with uneven follow-up times because it uses event ordering and risk sets rather than fixed time intervals.
4) Why were some rows dropped?
Rows are removed when time is nonpositive, event is not 0 or 1, a value is missing, or a covariate is not numeric.
5) What does the custom profile do?
It applies your fitted coefficients to a chosen covariate pattern. That creates a relative hazard estimate and a predicted survival curve.
6) Can I use categorical predictors?
Yes, but convert categories into numeric indicator columns first. For example, encode treatment groups as 0 and 1 dummy variables.
7) What does the C-index show?
The concordance index measures ranking ability. Higher values mean the model more often assigns greater risk to individuals who experience events earlier.
8) Why might convergence stop early?
Early stopping can happen with tiny samples, strong collinearity, sparse events, or very large covariate scales. Simplify inputs or rescale predictors.