Concept Architecture
Concept
Theoretically, the Cox?Snell Residual is a diagnostic measure used to assess the overall goodness-of-fit of a survival model. It is founded on survival analysis and residual diagnostics and exists to evaluate whether the fitted survival model adequately represents the observed time-to-event data. In health economics, Cox?Snell residuals are used to validate parametric and semi-parametric survival models that provide inputs for decision-analytic models and health technology assessments.
Mathematically, the Cox?Snell residual is defined as the estimated cumulative hazard evaluated at each individual's observed survival time. If the fitted model is correct, the residuals should follow an exponential distribution with a hazard of one. Model adequacy is commonly assessed by plotting the cumulative hazard of the residuals against the residual values, where an approximately 45-degree line indicates good fit.
In practice, Cox?Snell residuals are calculated after fitting survival models such as the Cox proportional hazards model or parametric survival models. They are routinely examined alongside other diagnostic measures to identify model misspecification, assess distributional assumptions and support selection of appropriate survival models for health economic evaluation.
Purpose
Used to assess the overall goodness-of-fit of survival models, identify model misspecification, compare alternative survival models and validate survival estimates used in health economic analyses.
Mathematical Formulae
Primary Formula
r? = H?(t?|X?)
where:
- r? = Cox?Snell residual
- H?(t?|X?) = estimated cumulative hazard at the observed survival time
Supporting Formulae
Cumulative hazard:
H(t) = ?ln(S(t))
Expected distribution under a correctly specified model:
r? ~ Exponential(1)
Related Mathematical Methods
- Survival model diagnostics
- Cumulative hazard estimation
- Cox proportional hazards model
- Parametric survival modelling
- Martingale residuals
- Deviance residuals
Example
A fitted Cox proportional hazards model estimates a patient's cumulative hazard at the observed event time as 0.85.
Cox?Snell residual:
r = 0.85
After calculating residuals for all patients, the cumulative hazard of the residuals is plotted against the residual values. If the points closely follow a straight line with slope 1 through the origin, the fitted survival model is considered to provide an adequate fit.
Excel Implementation
| Function | Example Formula | Health Economics Application |
|---|---|---|
| LN | =-LN(B2) | Calculate cumulative hazard from predicted survival probability. |
| EXP | =EXP(-B2) | Convert cumulative hazard into survival probability for diagnostic checks. |
| SCATTER Chart | Residuals vs cumulative hazard | Assess overall goodness-of-fit of the fitted survival model. |
VBA (Optional)
Automate calculation of Cox?Snell residuals and generate goodness-of-fit plots for survival model validation.
Sources
- Cox DR, Snell EJ. Analysis of Binary Data. Chapman and Hall.
- Collett D. Modelling Survival Data in Medical Research. CRC Press.
- Klein JP, Moeschberger ML. Survival Analysis: Techniques for Censored and Truncated Data. Springer.
- Therneau TM, Grambsch PM. Modeling Survival Data: Extending the Cox Model. Springer.
- Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation. Oxford University Press.
Related Concepts (2)
Library
Publications
1
NICE DSU Technical Support Document 15: Cost-effectiveness modelling using patient-level simulation — Davis, Stevenson, Tappenden & Wailoo, TSD 15 ed., 2014 (NICE Decision Support Unit (University of Sheffield))
Guidance on individual patient-level (microsimulation) cost-effectiveness modelling — when to use it in preference to cohort models, how to structure it, and how to handle the associated computational and uncertainty challenges.
Frequently Asked Questions (7)
What is a cox snell residual?
A residual used to check a survival model's overall fit, which should follow a standard exponential distribution if the fitted model is correct.
Source: Cox & Snell 1968
What is a Cox-Snell residual?
A Cox-Snell residual is a residual used to check the overall fit of a survival model, defined so that, if the fitted model is correct, the residuals should follow a standard exponential distribution. Proposed by Cox and Snell, these residuals transform each observation using the fitted model's cumulative hazard, and their agreement with the expected exponential distribution is examined to assess whether the model fits the data adequately. They provide a general diagnostic for survival model fit.
Source: Cox & Snell 1968
What goodness-of-fit diagnostic is a Cox-Snell residual?
A Cox-Snell residual is a diagnostic quantity used to assess how well a survival model fits the observed data overall. It is derived from the estimated cumulative hazard for each individual at their event or censoring time. If the model fits well, these residuals should behave like a sample from a standard exponential distribution, which can be checked graphically. Departures from that expected pattern signal that the chosen model does not describe the data adequately. The method gives a general check of fit rather than a test of any single assumption.
Source: Collett 2015
How are Cox-Snell residuals used?
Cox-Snell residuals are used by computing them from the fitted survival model, then checking whether they follow a standard exponential distribution, which they should if the model fits well. This is typically done by treating the residuals as a survival sample, estimating their cumulative hazard, and plotting it against the residuals: a straight line through the origin with slope one indicates good fit, while departures indicate poor fit. The comparison with the expected exponential distribution thus provides a graphical check of the model's overall adequacy.
Source: Cox & Snell 1968
Why should Cox-Snell residuals follow an exponential distribution?
Cox-Snell residuals should follow a standard exponential distribution when the model is correct because they are constructed from the model's cumulative hazard evaluated at each observation, and the cumulative hazard of the true survival distribution, evaluated at the event times, has this property. So if the fitted model matches the true one, the residuals inherit the standard exponential distribution. Departure from it indicates that the model does not fit, since the transformation would not produce exponential residuals if the model were wrong.
Source: Cox & Snell 1968
What do Cox-Snell residuals reveal about model fit?
Cox-Snell residuals reveal whether a survival model fits the data overall: agreement with the standard exponential distribution, shown by a straight cumulative-hazard plot with slope one, indicates good fit, while systematic departures indicate misfit, suggesting the model's form is inappropriate. They provide an overall diagnostic rather than pinpointing the cause of misfit, so they show whether the model fits adequately but may need to be supplemented by other residuals or checks to identify specific problems such as non-proportional hazards.
Source: Collett 2015
What are the limitations of Cox-Snell residuals?
Cox-Snell residuals assess overall model fit but do not identify the specific nature or source of any misfit, so a poor plot indicates a problem without showing whether it is due to the wrong distribution, non-proportional hazards, or omitted covariates. Their interpretation can be affected by censoring, and the plots can be hard to read near the tails. They are therefore used alongside other diagnostics that target particular assumptions, so Cox-Snell residuals provide a general check of fit rather than a complete diagnosis.
Source: Cox & Snell 1968
Trust Record
Verified by Dr Darrin Baines
British health economist
Professional identity: darrinbaines.org
Verification date: 20 Oct 2025
Content version: 1.0.0
Canonical Identity
- Persistent URI
- https://healtheconomics.wiki/concept/cox-snell-residual
- Term code
- HE-EM-SM-009
Stable URI · Machine-readable · Resolvable · CC BY 4.0