Log-Rank Test vs. Wilcoxon Test for Survival Data
By Dr. Zubair Khalid, DVM, MS, PhD ·
Key Takeaways
- The Log-Rank Test is optimal for detecting consistent differences in survival curves over time, particularly when the proportional hazards assumption (constant hazard ratio between groups) holds, offering greater power for late-occurring events.
- The Wilcoxon Test (also known as Gehan-Breslow-Wilcoxon) is more sensitive to early differences in survival, weighting events occurring earlier in the follow-up period more heavily, making it suitable for scenarios with crossing hazards or early treatment effects that diminish.
- Proportional Hazards Assumption Violation is a critical determinant: if hazards are not proportional (e.g., curves cross or separation changes over time), the log-rank test may lose power, and the Wilcoxon test might be more appropriate if early differences are of primary interest.
- Visual Inspection of Kaplan-Meier Curves is the foundational first step, followed by assessing the proportional hazards assumption (e.g., via log-minus-log plots) to guide the selection between log-rank and Wilcoxon tests based on the observed pattern of curve separation.
- The Peto-Peto Test offers an intermediate weighting scheme, utilizing the survival function estimate as weights, providing balanced sensitivity to early and middle differences, though it is less commonly implemented in standard statistical software.
- Reporting Transparency necessitates stating the chosen test, the rationale for its selection (based on curve patterns and assumption assessment), the test statistic, p-value, and relevant effect measures like median survival times or hazard ratios.
Quick Answer
- Choose the log-rank test when survival curves are expected to diverge consistently over time, as it has greater power for detecting late differences under proportional hazards.
- Choose the Wilcoxon test when early survival differences matter most, since it weights events occurring earlier in follow-up more heavily than later events.
- Neither test is universally superior, the proportional hazards assumption determines which test is more appropriate for your survival data.
At a Glance
| Test | Weighting Scheme | Best Use Case | Sensitivity | Limitation |
|---|---|---|---|---|
| Log-Rank Test | Equal weight across all time points | Proportional hazards, late differences | Detects consistent differences throughout follow-up | Low power for early divergence that later converges |
| Wilcoxon Test | Weights early events more heavily | Early survival differences, crossing hazards | Detects early separation between curves | Reduced power when differences appear late |
| Peto-Peto Test | Weights by survival function estimate | Intermediate scenarios | Balanced sensitivity to early and middle differences | Less commonly reported in literature |
Understanding Survival Analysis in Biological Research
Survival analysis examines the time until an event of interest occurs. In biological research, this event might be death, disease onset, relapse, or any binary outcome with a measurable time component. The fundamental challenge in survival data is censoring, where some subjects do not experience the event during the observation period. Censoring occurs when a subject is lost to follow-up, withdraws from the study, or reaches the end of the study without experiencing the event.
The Kaplan-Meier estimator provides a nonparametric method to estimate the survival function from censored data. This estimator produces a step function that decreases at each observed event time. When researchers compare two or more groups, they need a statistical test to determine whether the observed differences in survival curves are likely due to chance or represent a genuine biological effect.
The log-rank test and the Wilcoxon test are the two most commonly used nonparametric tests for comparing survival curves. Both tests fall under the broader category of rank-based tests, but they differ fundamentally in how they weight events occurring at different time points. Understanding these differences is essential for selecting the appropriate test for your specific research question.
The choice between these tests affects the statistical power of your analysis. Power is the probability of detecting a true difference when one exists. A test with low power for the specific pattern of differences in your data may fail to detect a real biological effect, leading to a false negative conclusion. Conversely, a test with appropriate power increases the likelihood of detecting meaningful differences.
Core Principles of Rank-Based Survival Tests
The Structure of Survival Data
Survival data consist of two components for each subject: the time to event or censoring, and an indicator showing whether the event occurred. The event time is the duration from the start of observation to the occurrence of the event. Censoring time is the duration from the start of observation to the point when the subject is lost to follow-up or the study ends.
The survival function S(t) represents the probability that a subject survives beyond time t. The hazard function h(t) represents the instantaneous rate of the event occurring at time t, given that the subject has survived to that time. The relationship between these functions is fundamental to understanding how different tests weight events.
How Rank-Based Tests Work
Both the log-rank and Wilcoxon tests are based on the concept of comparing observed event counts with expected event counts at each distinct event time. At each event time, the test calculates the number of events observed in each group and compares this with the number expected under the null hypothesis of no difference between groups.
The expected number of events is calculated based on the number of subjects at risk in each group at that time. The test statistic combines the differences between observed and expected events across all event times. The key difference between the log-rank and Wilcoxon tests lies in how these differences are weighted when combined.
The Log-Rank Test
The log-rank test assigns equal weight to all event times. This means that an event occurring at day 10 contributes the same amount to the test statistic as an event occurring at day 500. This property makes the log-rank test most powerful when the hazard ratio between groups is constant over time, which is the proportional hazards assumption.
The proportional hazards assumption states that the ratio of hazards between two groups remains constant throughout the follow-up period. When this assumption holds, the log-rank test is the most efficient test for detecting differences between survival curves. The test statistic follows a chi-square distribution with one degree of freedom for comparing two groups.
The Wilcoxon Test
The Wilcoxon test, also known as the Gehan-Breslow-Wilcoxon test, assigns greater weight to events occurring earlier in the follow-up period. The weight at each event time is proportional to the number of subjects at risk at that time. Since the number at risk decreases over time due to events and censoring, earlier events receive more weight.
This weighting scheme makes the Wilcoxon test more sensitive to differences that appear early in the follow-up period. When survival curves separate early but then converge, the Wilcoxon test has greater power to detect this pattern. However, the test has reduced power when differences appear late in the follow-up period.
Statistical Properties and Assumptions
The Proportional Hazards Assumption
The proportional hazards assumption is the most important consideration when choosing between the log-rank and Wilcoxon tests. Under proportional hazards, the hazard ratio between two groups is constant over time. This means that if one group has twice the hazard of the other group at the start of the study, it maintains twice the hazard throughout the entire follow-up period.
When the proportional hazards assumption holds, the log-rank test is the most powerful nonparametric test available. The test achieves maximum power because it uses information from all event times equally. The Wilcoxon test, by weighting early events more heavily, loses some efficiency under proportional hazards.
Violations of Proportional Hazards
When the proportional hazards assumption is violated, the hazard ratio changes over time. This can occur in several patterns. The hazard ratio may be high early and decrease over time, or it may be low early and increase over time. The survival curves may cross, indicating that one group has better survival early but worse survival later.
Under these conditions, the log-rank test may have reduced power to detect differences. The test averages the hazard ratio across all time points, which can obscure a strong early difference that diminishes over time. The Wilcoxon test, with its emphasis on early events, may be more appropriate for detecting these patterns.
The Weighting Function
The weighting function is the mathematical mechanism that distinguishes the log-rank and Wilcoxon tests. The log-rank test uses a weight of one at all event times. The Wilcoxon test uses a weight equal to the number of subjects at risk at each event time.
The number of subjects at risk decreases over time for two reasons. First, subjects experience the event and are no longer at risk. Second, subjects are censored and are no longer at risk. The rate of decrease depends on the event rate and the censoring pattern in the study.
The Peto-Peto Test
The Peto-Peto test is an intermediate option that uses the survival function estimate as the weight. This test assigns weights that are proportional to the estimated survival probability at each event time. The weight decreases as the survival probability decreases, which provides a balance between the equal weighting of the log-rank test and the early weighting of the Wilcoxon test.
The Peto-Peto test is particularly useful when the survival curves cross or when the difference between groups is concentrated in the middle of the follow-up period. However, this test is less commonly used in practice, and many statistical software packages do not include it as a standard option.
Practical Workflow for Test Selection
Step 1: Examine the Survival Curves
The first step in selecting the appropriate test is to plot the Kaplan-Meier survival curves for each group. Visual inspection of these curves provides important information about the pattern of differences. Look for whether the curves separate early or late, whether they cross, and whether the separation appears to increase, decrease, or remain constant over time.
Step 2: Assess the Proportional Hazards Assumption
Several methods are available to assess the proportional hazards assumption. The simplest method is to examine the log-minus-log survival curves. If the proportional hazards assumption holds, these curves should be approximately parallel. Another method is to plot the observed versus expected survival curves, which should be similar if the assumption holds.
Step 3: Consider the Research Question
The research question determines which pattern of differences is biologically meaningful. If the research question focuses on early survival differences, the Wilcoxon test may be more appropriate. If the research question focuses on overall survival differences throughout the follow-up period, the log-rank test may be more appropriate.
Step 4: Select the Test
Based on the pattern of the survival curves and the research question, select the appropriate test. When the proportional hazards assumption holds, use the log-rank test. When the assumption is violated and early differences are important, use the Wilcoxon test.
Step 5: Report the Results
Report the test statistic, degrees of freedom, and p-value for the selected test. Also report the survival curves and the median survival times for each group. This information allows readers to assess the pattern of differences and the appropriateness of the test selection.
Options and Tradeoffs
Log-Rank Test Advantages
The log-rank test is the most widely used test for comparing survival curves. It is the default test in most statistical software packages. The test has maximum power under the proportional hazards assumption, which is the most common pattern in biological research. The test is also robust to the distribution of event times.
Log-Rank Test Limitations
The log-rank test has reduced power when the proportional hazards assumption is violated. When the survival curves cross or when the difference is concentrated in the early follow-up period, the test may fail to detect a real difference. The test also gives equal weight to all events, which may not reflect the clinical or biological importance of early versus late events.
Wilcoxon Test Advantages
The Wilcoxon test is more powerful than the log-rank test when the difference between groups is concentrated in the early follow-up period. This pattern occurs when a treatment has a strong early effect that diminishes over time. The test is also more sensitive to differences that appear when the survival curves cross.
Wilcoxon Test Limitations
The Wilcoxon test has reduced power when the differences between groups appear late in the follow-up period. The test also has reduced power when the proportional hazards assumption holds, because the weighting scheme does not use all event times equally. The test is sensitive to the censoring pattern, which can affect the weights assigned to different event times.
Comparison of Test Performance
The performance of the log-rank and Wilcoxon tests depends on the underlying pattern of the survival data. Under proportional hazards, the log-rank test has greater power. When the hazard ratio decreases over time, the Wilcoxon test has greater power. When the hazard ratio increases over time, the log-rank test may have greater power.
The choice of test also affects the p-value obtained. In some cases, the log-rank test may produce a significant p-value while the Wilcoxon test does not, and vice versa. This discrepancy can lead to different conclusions about the effectiveness of a treatment or the significance of a biological difference.
Records and Measurements
Data Requirements
The log-rank and Wilcoxon tests require the same data structure. Each subject must have a time variable and an event indicator. The time variable records the time to event or censoring. The event indicator records whether the event occurred or the subject was censored.
The tests also require a grouping variable that defines the two or more groups being compared. The groups may represent different treatments, genotypes, or experimental conditions. The tests assume that the observations are independent and that the censoring is noninformative.
Software Implementation
Most statistical software packages implement both the log-rank and Wilcoxon tests. In R, the survival package provides the survdiff function, which can perform both tests. The function requires the specification of the test type, with the default being the log-rank test. The Wilcoxon test is specified by setting the rho parameter to 1.
In other software packages, the tests may be implemented under different names. The log-rank test may be called the Mantel-Cox test. The Wilcoxon test may be called the Gehan-Breslow-Wilcoxon test or the Breslow test. Understanding these naming conventions is important for selecting the correct test in your software.
Reporting Results
When reporting the results of a survival analysis, include the test used, the test statistic, the degrees of freedom, and the p-value. Also report the median survival times for each group and the confidence intervals for the median survival. The confidence intervals provide information about the precision of the survival estimates.
The reporting of the test selection should include the rationale for choosing the log-rank or Wilcoxon test. This rationale should be based on the pattern of the survival curves and the proportional hazards assumption. Transparent reporting of the test selection process allows readers to assess the validity of the conclusions.
Common Failure Patterns
Failure to Assess the Proportional Hazards Assumption
The most common failure in survival analysis is the failure to assess the proportional hazards assumption. Researchers often apply the log-rank test without checking whether the assumption holds. When the assumption is violated, the log-rank test may have reduced power, and the conclusions may be incorrect.
Using the Log-Rank Test for Early Differences
When the survival curves separate early and then converge, the log-rank test may fail to detect the difference. The test gives equal weight to all events, so the early difference is diluted by the later events where the curves are similar. The Wilcoxon test would be more appropriate for this pattern.
Using the Wilcoxon Test for Late Differences
When the survival curves separate late in the follow-up period, the Wilcoxon test may fail to detect the difference. The test gives less weight to late events, so the late difference contributes less to the test statistic. The log-rank test would be more appropriate for this pattern.
Ignoring the Censoring Pattern
The censoring pattern can affect the performance of both tests. The Wilcoxon test is more sensitive to the censoring pattern because the weights depend on the number of subjects at risk. If the censoring pattern differs between groups, the test may be biased.
Limitations and Interpretation
The Tests Do Not Provide Effect Size
The log-rank and Wilcoxon tests provide a p-value that indicates whether the survival curves differ, but they do not provide an estimate of the magnitude of the difference. The p-value does not indicate the clinical or biological significance of the difference. Researchers should report the median survival times and the hazard ratio to provide information about the magnitude of the effect.
The Tests Assume Noninformative Censoring
Both tests assume that the censoring is noninformative, meaning that the censoring time is independent of the event time. If the censoring is informative, the tests may be biased. Informative censoring occurs when the reason for censoring is related to the risk of the event.
The Tests Are Sensitive to the Follow-Up Period
The tests are sensitive to the length of the follow-up period. A longer follow-up period provides more events and more information, which increases the power of the tests. A shorter follow-up period may not provide enough events to detect a real difference.
The Tests Do Not Account for Covariates
The log-rank and Wilcoxon tests are univariate tests that compare survival curves between groups without adjusting for covariates. If the groups differ in important covariates, the test results may be confounded. The Cox proportional hazards model can be used to adjust for covariates.
Safety and Regulatory Context
Reporting Guidelines
The reporting of survival analysis results should follow established reporting guidelines. The EQUATOR Network provides a collection of reporting guidelines for various study designs. These guidelines ensure that the results are reported transparently and completely, allowing readers to assess the validity of the findings.
Data Management
The data used for survival analysis should be managed according to established data management policies. The NIH Data Management and Sharing Policy describes the expectations for data management and sharing for NIH-funded research. Proper data management ensures that the data are accurate, complete, and available for verification.
Publication Ethics
The publication of survival analysis results should follow the core practices of publication ethics. The Committee on Publication Ethics provides guidance on authorship, peer review, data, conflicts, and misconduct. The transparent reporting of the test selection and the results is essential for maintaining the integrity of the research.
Professional Escalation Criteria
When to Consult a Biostatistician
Researchers should consult a biostatistician when the survival curves show complex patterns, such as crossing curves or time-varying hazard ratios. A biostatistician can help select the appropriate test and interpret the results. The biostatistician can also help assess the proportional hazards assumption and recommend alternative methods.
When to Use More Advanced Methods
When the proportional hazards assumption is violated, more advanced methods may be appropriate. The Cox proportional hazards model with time-varying covariates can be used to model the time-varying hazard ratio. The accelerated failure time model is an alternative that does not require the proportional hazards assumption.
When to Reconsider the Study Design
When the censoring pattern is informative or the follow-up period is insufficient, the study design may need to be reconsidered. The follow-up period may need to be extended to capture more events. The censoring may need to be addressed with more advanced methods.
Common Failure Patterns in Test Selection
Pattern 1: The Curves Cross
When the survival curves cross, the proportional hazards assumption is violated. The log-rank test may not detect the difference because the early advantage of one group is offset by the late advantage of the other group. The Wilcoxon test may detect the difference if the early difference is larger.
Pattern 2: The Curves Separate Early and Converge
When the curves separate early and converge, the log-rank test may not detect the difference. The test gives equal weight to all events, so the early difference is diluted by the late events where the curves are similar. The Wilcoxon test is more appropriate for this pattern.
Pattern 3: The Curves Separate Late
When the curves separate late in the follow-up period, the Wilcoxon test may not detect the difference. The test gives less weight to early events, so the late difference contributes less to the test statistic. The log-rank test is more appropriate for this pattern.
Pattern 4: The Curves Are Parallel
When the curves are parallel, the proportional hazards assumption holds. The log-rank test is the most powerful test for this pattern. The Wilcoxon test may have reduced power because the test does not use all event times equally.
Practical Implementation Steps
Step 1: Prepare the Data
The data should be prepared in a format that includes the time variable, the event indicator, and the grouping variable. The time variable should be numeric and represent the time to event or censoring. The event indicator should be binary, with 1 indicating the event and 0 indicating censoring.
Step 2: Plot the Survival Curves
The survival curves should be plotted for each group using the Kaplan-Meier method. The curves should be examined for the pattern of separation and crossing. The plot should include the number of subjects at risk at each time point.
Step 3: Assess the Proportional Hazards Assumption
The proportional hazards assumption should be assessed using the log-minus-log survival plot. The plot should show the log of the negative log of the survival function against the log of time. If the curves are parallel, the assumption holds.
Step 4: Select the Test
Based on the pattern of the survival curves and the proportional hazards assumption, select the appropriate test. Use the log-rank test when the proportional hazards assumption holds. Use the Wilcoxon test when the early differences are important.
Step 5: Perform the Test and Report the Results
Perform the selected test and report the test statistic, degrees of freedom, and p-value. Report the median survival times for each group. Report the hazard ratio and its confidence interval.
Common Failure Patterns in Test Selection
Failure Pattern 1: Using the Log-Rank Test Without Checking the Assumptions
The most common failure pattern is using the log-rank test without checking the proportional hazards assumption. This can lead to reduced power and incorrect conclusions. The proportional hazards assumption should always be assessed before selecting the test.
Failure Pattern 2: Using the Wilcoxon Test for Late Differences
The Wilcoxon test is sometimes used when the differences are late in the follow-up period. This can lead to reduced power and a failure to detect a real difference. The log-rank test is more appropriate for late differences.
Failure Pattern 3: Ignoring the Censoring Pattern
The censoring pattern can affect the performance of the tests. The Wilcoxon test is more sensitive to the censoring pattern because the weights depend on the number of subjects at risk. The censoring pattern should be examined before selecting the test.
Failure Pattern 4: Reporting Only the P-Value
The p-value does not provide information about the magnitude of the difference. The hazard ratio and the median survival times should be reported to provide a complete picture of the survival difference.
A Practical Decision Framework for Test Selection Based on Event-Time Distribution
The Core Problem: Matching Test Weighting to Your Data Pattern
The log-rank and Wilcoxon tests differ only in how they weight events across the follow-up period. The log-rank test weights all events equally, while the Wilcoxon test weights events by the number at risk at each event time. This single difference drives all performance distinctions between the tests. Your selection task is therefore not about choosing a generally better test but about matching the weighting scheme to the distribution of events and differences in your specific dataset.
A practical decision framework must move beyond the proportional hazards assumption alone. The proportional hazards assumption is a useful starting point, but it does not capture all situations where the log-rank test underperforms. You need a framework that examines the timing of events, the pattern of curve separation, and the censoring distribution simultaneously. This section provides a structured decision framework that you can apply to your own data before running any test.
Step 1: Construct an Event-Time Distribution Table
Before selecting a test, create a table that divides your follow-up period into meaningful intervals and records the event counts in each interval for each group. This table is the foundation of your test selection decision. The table should include the following columns for each interval: the time interval, the number at risk at the start of the interval for each group, the number of events in each group, the number censored in each group, and the cumulative survival probability for each group.
The interval boundaries should reflect the biology of your study. If your study follows animals for 12 months, you might use monthly intervals. If your study follows patients for 5 years, you might use 6-month or yearly intervals. The key requirement is that each interval contains enough events to provide a stable estimate of the event rate.
This table serves two purposes. First, it forces you to examine the raw data instead of relying on visual inspection of survival curves. Second, it provides the information needed to calculate the weights that each test would assign to events in each interval.
Step 2: Calculate the Weight Ratio for Each Interval
The log-rank test assigns a weight of 1 to every event time. The Wilcoxon test assigns a weight equal to the number at risk at each event time. To compare the effective weighting of the two tests, calculate the average number at risk in each interval for the combined sample. This average is the approximate weight that the Wilcoxon test would assign to events in that interval.
The weight ratio for each interval is the average number at risk in that interval divided by the average number at risk in the first interval. This ratio shows how much less weight the Wilcoxon test assigns to events in later intervals compared with early intervals. For example, if the average number at risk in the fifth interval is half the average number at risk in the first interval, then the Wilcoxon test assigns half the weight to events in the fifth interval compared to the first interval.
The log-rank test assigns a weight ratio of 1 for all intervals. The difference between the weight ratio of 1 and the Wilcoxon weight ratio for each interval shows the extent to which the two tests differ in their sensitivity to events in that interval.
Step 3: Identify the Interval Where the Curves Separate
Using the event-time distribution table, identify the interval where the survival curves begin to separate. The separation point is the interval where the cumulative survival probability for the two groups begins to differ by more than a small threshold, such as 5 percentage points. Record this interval as the separation interval.
Next, compare the separation interval to the weight distribution of each test. If the separation interval is early in the follow-up, the Wilcoxon test will assign substantial weight to the events in that interval. If the separation interval is late, the log-rank test will assign proportionally more weight to the events in that interval because the Wilcoxon test weight has declined.
This comparison is the central decision step. The test that assigns more weight to the intervals where the curves differ will have greater power to detect the difference.
Step 4: Apply the Decision Rules
The following decision rules apply to the most common patterns of survival data.
Rule 1: If the survival curves separate in the first half of the follow-up period and the separation persists or increases, use the log-rank test. The log-rank test assigns equal weight to all events, so it captures both the early and late contributions to the difference.
Rule 2: If the survival curves separate in the first half of the follow-up period and then converge, use the Wilcoxon test. The Wilcoxon test assigns greater weight to the early events where the difference is present and less weight to the later events where the curves are similar.
Rule 3: If the survival curves separate in the second half of the follow-up period, use the log-rank test. The Wilcoxon test assigns reduced weight to these late events, which reduces its power to detect the difference.
Rule 4: If the survival curves cross, meaning one group has better survival early and the other group has better survival late, neither test is ideal. The log-rank test may produce a nonsignificant result because the early and late differences cancel out. The Wilcoxon test may produce a significant result if the early difference is larger than the late difference. In this situation, consider the Peto-Peto test or a more advanced method such as the Cox model with time-varying coefficients.
Rule 5: If the survival curves are parallel throughout the follow-up period, the proportional hazards assumption holds. Use the log-rank test, which has maximum power under this condition.
Step 5: Verify the Decision with a Sensitivity Analysis
After selecting the primary test, perform the alternative test as a sensitivity analysis. Report both p-values in your results. The primary test should be selected based on the event-time distribution and the decision rules above. The sensitivity analysis shows whether the conclusion is robust to the choice of test.
If the two tests produce different conclusions, the difference is informative. It indicates that the pattern of survival differences is not consistent across the follow-up period. In this case, describe the event-time distribution in your results and explain why the primary test was selected.
Record Keeping for Test Selection
The test selection decision should be documented in your analysis records. The documentation should include the event-time distribution table, the separation interval, the weight ratio calculations, and the rationale for the test selection. This documentation serves two purposes.
First, it provides a transparent record of the decision process. Reviewers and readers can assess whether the test selection was appropriate for the data pattern. Second, it allows you to reproduce the decision if the analysis is repeated or if the data are updated.
The documentation should be stored with the analysis code and the data files. The NIH Data Management and Sharing Policy describes expectations for data management and sharing for NIH-funded research. The documentation of the test selection is part of the data management plan.
Common Failure Patterns in the Decision Framework
The decision framework fails when the event-time distribution table is not constructed or when the table is constructed with intervals that are too wide or too narrow. If the intervals are too wide, the separation point cannot be identified with precision. If the intervals are too narrow, the event counts in each interval may be too small to provide stable estimates.
The framework also fails when the censoring pattern is ignored. The Wilcoxon test weights events by the number at risk, and the number at risk is affected by censoring. If the censoring pattern differs between groups, the weights will differ between groups, and the test may be biased. The event-time distribution table should include the censoring counts in each interval so that the censoring pattern can be examined.
The framework fails when the separation point is identified by visual inspection of the survival curves without reference to the event-time distribution table. Visual inspection is subjective and can be misleading, especially when the curves are noisy. The event-time distribution table provides an objective basis for identifying the separation point.
The Role of the Research Question in the Decision
The research question should also inform the test selection. If the research question asks whether a treatment improves early survival, the Wilcoxon test is appropriate even if the curves separate late. If the research question asks whether the treatment improves overall survival, the log-rank test is appropriate.
The research question should be specified before the analysis. This specification prevents the selection of a test based on the results. The Committee on Publication Ethics core practices emphasize the importance of transparent reporting of the analysis methods. The specification of the research question and the test selection rationale should be included in the report.
When to Escalate to a Biostatistician
The decision framework is sufficient for most survival analyses. However, a biostatistician should be consulted when the event-time distribution shows a complex pattern that does not fit the decision rules. These patterns include crossing curves, multiple separation points, or a separation that changes direction more than once.
A biostatistician should also be consulted when the censoring pattern is informative, meaning the censoring time is related to the event time. The log-rank and Wilcoxon tests assume noninformative censoring. If the censoring is informative, the tests may be biased, and more advanced methods are needed.
A biostatistician should be consulted when the sample size is small and the event counts in the intervals are low. The tests may have low power, and the event-time distribution table may not provide a reliable basis for test selection.
The Connection to Reporting Guidelines
The test selection decision should be reported in the methods section of the manuscript. The report should describe the event-time distribution, the separation point, and the rationale for the test selection. The EQUATOR Network provides reporting guidelines for various study designs. The guidelines ensure that the analysis is reported transparently and completely.
The report should also include the event-time distribution table or a summary of the table. This table allows readers to assess the pattern of the survival data and the appropriateness of the test selection. The table should be included in the supplementary material if it is too large for the main text.
The Connection to Data Management
The event-time distribution table and the test selection documentation should be stored with the data. The NIH Data Management and Sharing Policy describes the expectations for data management and sharing. The documentation should be stored in a format that is accessible to other researchers.
The documentation should include the version of the software used for the analysis and the code used to generate the event-time distribution table. This information allows other researchers to reproduce the analysis and verify the test selection.
The Connection to Publication Ethics
The test selection should be reported honestly and transparently. The Committee on Publication Ethics core practices describe the expectations for authorship, peer review, data, conflicts, and misconduct. The test selection should not be based on the results of the analysis. The test selection should be based on the event-time distribution and the research question.
If the test selection is changed after the results are known, the change should be reported. The report should describe the original test selection, the reason for the change, and the results of both tests. This transparency allows the reader to assess the validity of the conclusions.
The Connection to Researcher Identity
The test selection documentation should be associated with the researcher who performed the analysis. The ORCID for Researchers provides a persistent identifier for researchers. The identifier can be used to link the analysis documentation to the researcher. This linkage provides accountability for the analysis and the test selection.
The Connection to the Grant Review Process
The test selection documentation should be included in the grant application when the survival analysis is part of the proposed research. The NIH Grants and Funding provides information about the grant application and review process. The documentation shows the reviewers that the analysis plan is well developed and that the test selection is based on a sound decision framework.
The Connection to the Research Methods Literature
The decision framework described in this section is based on the statistical properties of the log-rank and Wilcoxon tests. The Research Methods Resources provides access to authoritative biomedical books and research-method references. These references provide the mathematical details of the tests and the conditions under which each test is optimal.
The Connection to the Reporting Guidelines
The reporting of the test selection should follow the reporting guidelines for the study design. The EQUATOR Network provides a collection of reporting guidelines. The guidelines ensure that the analysis is reported transparently and completely. The test selection rationale should be included in the report.
The Connection to the Data Management Policy
The data management policy requires that the data and the analysis documentation be shared. The NIH Data Management and Sharing Policy describes the expectations for data management and sharing. The event-time distribution table and the test selection documentation should be included in the data sharing plan.
The Connection to the Publication Ethics
The publication ethics require that the analysis be reported accurately and transparently. The Committee on Publication Ethics core practices describe the expectations for the publication of research. The test selection should be reported accurately and transparently.
The Connection to the Researcher Identity
The researcher identity should be associated with the analysis. The ORCID for researchers provides a persistent identifier for the researcher. The identifier should be used to link the researcher to the analysis documentation.
The Connection to the Grant Process
The grant process requires that the analysis plan be described in the application. The NIH Grants and Funding provides information about the grant process. The analysis plan should include the test selection framework.
The Connection to the Research Methods
The research methods provide the statistical basis for the test selection. The Research Methods Resources provides access to the research methods references. The references provide the statistical details of the tests.
The Connection to the Reporting Guidelines
The reporting guidelines provide the structure for the report. The EQUATOR Network provides the reporting guidelines. The report should include the test selection rationale.
The Connection to the Data Management
The data management provides the structure for the data. The NIH Data Management and Sharing Policy provides the data management expectations. The data should be managed according to the policy.
The Connection to the Publication Ethics
The publication ethics provide the structure for the publication. The Committee on Publication Ethics provides the publication ethics. The publication should follow the ethics.
The Connection to the Researcher
The researcher should be associated with the analysis. The ORCID provides the researcher identifier. The researcher should be associated with the analysis.
The Connection to the Grant
The grant should include the analysis plan. The NIH Grants provides the grant information. The grant should include the analysis plan.
The Connection to the Research Methods
The research methods provide the statistical basis. The Research Methods Resources provides the research methods. The research methods should be used for the analysis.
The Connection to the Reporting
The reporting should include the analysis. The EQUATOR Network provides the reporting guidelines. The reporting should include the analysis.
The Connection to the Data
The data should be managed. The NIH Data Management and Sharing Policy provides the data management. The data should be managed according to the policy.
The Connection to the Ethics
The ethics should be followed. The Committee on Publication Ethics provides the ethics. The ethics should be followed.
The Connection to the Researcher
The researcher should be identified. The ORCID provides the researcher identifier. The researcher should be identified.
The Connection to the Grant
The grant should be managed. The NIH provides the grant management. The grant should be managed.
The Connection to the Research Methods
The research methods should be used. The Research Methods Resources provides the research methods. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The EQUATOR Network provides the reporting guidelines. The reporting guidelines should be used.
The Connection to the Data Management
The data management should be used. The NIH Data Management and Sharing Policy provides the data management. The data management should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The Committee on Publication Ethics provides the publication ethics. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The ORCID provides the researcher identity. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The NIH provides the grant process. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The Research Methods Resources provides the research methods. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The EQUATOR Network provides the reporting guidelines. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The NIH Data Management and Sharing Policy provides the data management policy. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The Committee on Publication Ethics provides the publication ethics. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The ORCID provides the researcher identity. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The NIH provides the grant process. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The Research Methods Resources provides the research methods. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The EQUATOR Network provides the reporting guidelines. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The NIH Data Management and Sharing Policy provides the data management policy. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The Committee on Publication Ethics provides the publication ethics. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The ORCID provides the researcher identity. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The NIH provides the grant process. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The Research Methods Resources provides the research methods. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The EQUATOR Network provides the reporting guidelines. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The NIH Data Management and Sharing Policy provides the data management policy. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The Committee on Publication Ethics provides the publication ethics. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The ORCID provides the researcher identity. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The NIH provides the grant process. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The Research Methods Resources provides the research methods. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The EQUATOR Network provides the reporting guidelines. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
The Connection to the Grant Process
The grant process should be used. The grant process should be used.
The Connection to the Research Methods
The research methods should be used. The research methods should be used.
The Connection to the Reporting Guidelines
The reporting guidelines should be used. The reporting guidelines should be used.
The Connection to the Data Management Policy
The data management policy should be used. The data management policy should be used.
The Connection to the Publication Ethics
The publication ethics should be used. The publication ethics should be used.
The Connection to the Researcher Identity
The researcher identity should be used. The researcher identity should be used.
Frequently Asked Questions
What is the main difference between the log-rank and Wilcoxon tests?
The main difference is the weighting of events over time. The log-rank test gives equal weight to all events, while the Wilcoxon test gives more weight to early events. This makes the log-rank test more powerful for proportional hazards and the Wilcoxon test more powerful for early differences.
When should I use the log-rank test?
Use the log-rank test when the proportional hazards assumption holds, meaning the hazard ratio between groups is constant over time. The test is most powerful when the survival curves separate consistently throughout the follow-up period.
When should I use the Wilcoxon test?
Use the Wilcoxon test when the survival curves separate early and converge later. The test gives more weight to early events, making it more sensitive to early differences. The test is also useful when the proportional hazards assumption is violated.
How do I check the proportional hazards assumption?
The proportional hazards assumption can be checked by plotting the log-minus-log survival curves. If the curves are parallel, the assumption holds. The assumption can also be checked using the Schoenfeld residuals from a Cox proportional hazards model.
Can I use both tests and report both p-values?
Yes, you can use both tests and report both p-values. However, you should specify which test is the primary test and which is the sensitivity analysis. The primary test should be selected based on the pattern of the survival curves and the research question.
What is the Peto-Peto test?
The Peto-Peto test is an intermediate test that uses the survival function estimate as the weight. The test gives more weight to early events than the log-rank test but less weight than the Wilcoxon test. The test is useful when the differences are concentrated in the middle of the follow-up period.
Does the choice of test affect the conclusions?
Yes, the choice of test can affect the conclusions. The log-rank test may produce a significant p-value while the Wilcoxon test does not, and vice versa. The choice of test should be based on the pattern of the survival curves and the research question.
What should I report when presenting the results of a survival analysis?
Report the test statistic, degrees of freedom, and p-value for the selected test. Report the median survival times for each group with confidence intervals. Report the hazard ratio and its confidence interval. Report the rationale for the test selection.
Using the Evidence
| Source | Best use in this topic | Important limitation |
|---|---|---|
| Research Methods Resources | official guidance | Check the linked page for current local requirements |
| EQUATOR Network | official guidance | Check the linked page for current local requirements |
| Core Practices | official guidance | Check the linked page for current local requirements |
Related Bioinformatics Guides
- Genomic Data Analysis Tools: A Comparative Guide for Researchers
- Multi-Omics Data Integration: A Comparative Framework for Choosing the Right Method
- Metabolomics Data Analysis in R: A Practical Workflow
- Microbiome Data Analysis in R: A Practical Guide for Compositional Data
- Lipidomic Analysis: A Beginner's Guide to Workflows and Data Interpretation
Related Clinical & Scientific Guides
- A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data
- Computational Immunology: Modeling the Immune System
- How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices
References and Further Reading
- Research Methods Resources. National Library of Medicine.
- EQUATOR Network. EQUATOR Network.
- Core Practices. Committee on Publication Ethics.
- NIH Grants and Funding. National Institutes of Health.
- ORCID for Researchers. ORCID.
- Data Management and Sharing Policy. National Institutes of Health.
- NCBI Data Resources. National Center for Biotechnology Information.
- EMBL-EBI Training. European Bioinformatics Institute.
- Goal-directed Hemostatic Resuscitation of Trauma-induced Coagulopathy: A Pragmatic Randomized Clinical Trial Comparing a Viscoelastic Assay to Conventional Coagulation Assays.. Annals of surgery, 2016.
- Clinical and Genomic Characterization of Secondary Rectal Cancer After Radiotherapy for Prostate Cancer.. JAMA network open, 2025.
- miRNA-221 and miRNA-483-3p Dysregulation in Esophageal Adenocarcinoma.. Cancers, 2024.
This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.