This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
CREDO 2 was a randomized, double-blind, parallel-group phase 3 trial evaluating two dosing regimens of olokizumab (OKZ) in subjects with active rheumatoid arthritis who were taking methotrexate. The trial included placebo and adalimumab comparator groups and evaluated binary and continuous efficacy outcomes at prespecified time points.
| Feature | CREDO 2 |
|---|---|
| Phase | Phase 3 |
| Condition | Rheumatoid arthritis |
| Design | Randomized, double-blind, parallel-group |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 1648 |
| Interventions | Olokizumab 64 mg q4w; olokizumab 64 mg q2w; adalimumab 40 mg q2w; placebo q2w |
| Primary endpoint | Percentage of subjects achieving ACR20 response at Week 12 |
| Results posted | Yes |
| Statistical analyses posted | 19 |
| ClinicalTrials.gov | NCT02760407 |
2. Clinical Question
The primary efficacy question was whether either olokizumab regimen produced a higher percentage of ACR20 responders at Week 12 than placebo in subjects with rheumatoid arthritis who were taking methotrexate but had active disease. The trial also compared olokizumab with adalimumab for ACR20 response and evaluated other measures of disease activity and physical ability.
Population
Subjects with rheumatoid arthritis who were taking methotrexate but had active disease.
Intervention
Olokizumab 64 mg every 4 weeks or 64 mg every 2 weeks, with methotrexate.
Comparator
Placebo every 2 weeks with methotrexate; adalimumab 40 mg every 2 weeks with methotrexate was also evaluated as an active comparator.
Primary question
Does either olokizumab regimen increase the Week 12 ACR20 response rate relative to placebo?
3. Trial Design
Olokizumab 64 mg q4w + Methotrexate
- Olokizumab 64 mg every 4 weeks
- Methotrexate
- ACR20 primary comparison versus placebo at Week 12
Olokizumab 64 mg q2w + Methotrexate
- Olokizumab 64 mg every 2 weeks
- Methotrexate
- ACR20 primary comparison versus placebo at Week 12
Adalimumab 40 mg q2w + Methotrexate
- Adalimumab 40 mg every 2 weeks
- Methotrexate
- Active comparator for ACR20 and other efficacy outcomes
Placebo q2w + Methotrexate
- Placebo every 2 weeks
- Methotrexate
- Placebo comparator for primary efficacy analyses
4. Primary Endpoint
The registered primary endpoint was the percentage of subjects achieving an American College of Rheumatology 20% (ACR20) response at Week 12. The registry defines an ACR20 responder as a subject satisfying the ACR20 criteria and remaining on randomized treatment and in the study at Week 12.
The registry therefore combines an efficacy response definition with a treatment/study-retention requirement. This is important statistically because the binary endpoint is not simply a measurement of symptom improvement: the subject must also remain on randomized treatment and in the study at the Week 12 assessment.
5. Statistical Methodology
Intent-to-treat analysis
The primary analysis population was the intent-to-treat (ITT) population. The registry defines this as all randomized subjects, with subjects analyzed according to the treatment group to which they were randomized. This preserves the treatment assignment created by randomization for the primary efficacy comparison.
Chi-squared testing for binary endpoints
The primary ACR20 comparisons used a 2×2 chi-squared framework. Each subject contributes to one of two categories—ACR20 responder or non-responder—and the treatment groups are compared through their resulting response proportions.
A positive risk difference indicates a higher response probability in the first-listed treatment group. An RD of 0.270 corresponds to a 27.0-percentage-point difference because the effect measure is expressed on the probability scale.
ANCOVA for HAQ-DI
The Week 12 analysis of improvement in physical ability, measured by HAQ-DI, used ANCOVA. The effect measure was the least squares mean difference between treatment groups.
The registry specifies that subjects with a missing baseline were excluded from this analysis. Data after treatment discontinuation were included, while data after discontinuing the study were multiply imputed based on the return-to-baseline assumption.
Non-inferiority comparisons
For comparisons between olokizumab and adalimumab, the registry reports non-inferiority hypotheses for ACR20 and low disease activity. The key quantity is the lower confidence-limit relative to the prespecified non-inferiority margin, rather than whether a conventional superiority p-value is below a threshold.
The registry states that these margins were used for the corresponding olokizumab-versus-adalimumab comparisons.
6. Primary Results: ACR20 at Week 12
The registry reports two primary analyses, one for each olokizumab dosing regimen versus placebo. Both comparisons used the ITT population, a chi-squared test, and risk difference as the effect measure.
Olokizumab 64 mg q4w vs Placebo
Risk difference in ACR20 response
97.5% two-sided CI: 0.183–0.352 · P < 0.0001
Analysis population: ITT · Chi-squared test · Superiority
The risk difference of 0.270 means that the Week 12 ACR20 response probability was estimated to be 0.270 higher in the olokizumab 64 mg q4w group than in the placebo q2w group. On a percentage-point scale, that is a 27.0-percentage-point difference.
The estimate is a difference in response probabilities, not a relative risk or odds ratio. It does not mean that an individual subject has a 27.0% greater personal probability of response, nor does it describe the magnitude of improvement among subjects who responded.
The 97.5% confidence interval of 0.183 to 0.352 describes uncertainty around the estimated risk difference under the analysis framework. Its width shows that the exact magnitude of the treatment-group difference is less certain than the direction of the estimated difference.
The p-value of <0.0001 addresses the evidence against the null hypothesis of no difference under the specified test. It does not measure the size, clinical importance, or precision of the treatment effect; those questions are addressed by the risk difference and confidence interval.
Olokizumab 64 mg q2w vs Placebo
Risk difference in ACR20 response
97.5% two-sided CI: 0.171–0.341 · P < 0.0001
Analysis population: ITT · Chi-squared test · Superiority
The risk difference of 0.258 corresponds to an estimated 25.8-percentage-point difference in Week 12 ACR20 response between olokizumab 64 mg q2w and placebo q2w.
The 97.5% confidence interval extends from 0.171 to 0.341. This interval describes uncertainty around the estimated difference; it does not describe the range of effects for individual subjects.
The p-value of <0.0001 indicates strong statistical evidence against a null difference under the reported chi-squared test. It should not be interpreted as a measure of treatment-effect magnitude. The effect magnitude is represented by the risk difference, while the confidence interval conveys its statistical precision.
As with the q4w analysis, the comparison is based on randomized treatment assignment in the ITT population. The binary endpoint also incorporates remaining on randomized treatment and in the study at Week 12, as specified in the endpoint definition.
| Primary comparison | Effect measure | Estimate | 97.5% two-sided CI | P-value |
|---|---|---|---|---|
| Olokizumab 64 mg q4w + methotrexate vs placebo q2w + methotrexate | Risk difference | 0.270 | 0.183–0.352 | <0.0001 |
| Olokizumab 64 mg q2w + methotrexate vs placebo q2w + methotrexate | Risk difference | 0.258 | 0.171–0.341 | <0.0001 |
7. ACR20 Comparison With Adalimumab
The registry also reports ACR20 comparisons involving the active comparator. These analyses are particularly useful for understanding the trial's non-inferiority framework because the treatment effect is expressed as a risk difference for olokizumab relative to adalimumab.
| Comparison | Estimate | 97.5% two-sided CI | Hypothesis | Registry criterion |
|---|---|---|---|---|
| Olokizumab q4w vs adalimumab q2w | 0.045 | -0.022 to 0.112 | Non-inferiority | Lower limit > -12% |
| Olokizumab q2w vs adalimumab q2w | 0.034 | -0.035 to 0.102 | Non-inferiority | Lower limit > -12% |
For the q4w comparison, the estimated risk difference was 0.045, with a 97.5% confidence interval from -0.022 to 0.112. For the q2w comparison, the estimate was 0.034, with a 97.5% confidence interval from -0.035 to 0.102.
These confidence intervals include zero, so the estimates do not establish superiority over adalimumab. That is not the question posed by the non-inferiority framework. The registry specifies a -12% non-inferiority margin: non-inferiority is achieved if the lower confidence limit is greater than that margin.
This illustrates why non-inferiority should not be reduced to the question of whether a conventional p-value is below 0.05. The relevant comparison is between the confidence interval and the prespecified margin.
8. Secondary Results: Low Disease Activity at Week 12
Low disease activity was another binary Week 12 endpoint. The registry reports chi-squared analyses for comparisons against placebo and non-inferiority analyses for comparisons with adalimumab.
| Comparison | Estimate | CI | P-value | Hypothesis |
|---|---|---|---|---|
| Olokizumab q4w vs placebo | 0.332 | 97.5%: 0.257–0.397 | <0.0001 | Superiority |
| Olokizumab q2w vs placebo | 0.325 | 97.5%: 0.250–0.391 | <0.0001 | Superiority |
| Adalimumab q2w vs placebo | 0.256 | 95%: 0.191–0.313 | <0.0001 | Superiority |
| Olokizumab q4w vs adalimumab q2w | 0.076 | 97.5%: 0.004–0.147 | Not reported | Non-inferiority |
| Olokizumab q2w vs adalimumab q2w | 0.069 | 97.5%: -0.003–0.141 | Not reported | Non-inferiority |
The placebo comparisons show positive risk differences of 0.332 and 0.325 for the two olokizumab regimens, respectively. These correspond to estimated 33.2- and 32.5-percentage-point differences in the probability of achieving low disease activity at Week 12.
The active-comparator estimates are interpreted differently. The q4w versus adalimumab estimate of 0.076 has a lower 97.5% confidence limit of 0.004, which is above the registry-specified non-inferiority margin of -7.5%. The q2w estimate of 0.069 has a lower limit of -0.003, also above -7.5%.
These are non-inferiority comparisons, not superiority tests. The absence of a reported p-value for these analyses is therefore not a missing ingredient for interpreting the stated non-inferiority criterion; the confidence interval relative to the margin is the key quantity.
9. Secondary Results: HAQ-DI at Week 12
The registry reports improvement in physical ability from baseline to Week 12 using the Health Assessment Questionnaire-Disability Index (HAQ-DI). These analyses used ANCOVA and reported least squares mean differences.
| Comparison | Least squares mean difference | CI | P-value |
|---|---|---|---|
| Olokizumab q4w vs placebo | -0.19 | 97.5%: -0.29 to -0.09 | <0.0001 |
| Olokizumab q2w vs placebo | -0.22 | 97.5%: -0.33 to -0.12 | <0.0001 |
| Adalimumab q2w vs placebo | -0.19 | 95%: -0.28 to -0.10 | Not reported |
The negative least squares mean differences indicate lower adjusted HAQ-DI values in the first-listed active-treatment group relative to placebo. For olokizumab, the estimated differences were -0.19 for q4w and -0.22 for q2w.
The confidence intervals exclude zero for both reported olokizumab-versus-placebo analyses. For q4w, the 97.5% interval was -0.29 to -0.09; for q2w it was -0.33 to -0.12.
ANCOVA provides a model-based adjusted comparison rather than simply subtracting two observed group means. The registry also specifies the handling of missing baseline and post-discontinuation observations, which is important because the estimated treatment difference depends partly on the analysis population and missing-data strategy.
10. Secondary Results: ACR50 at Week 24
ACR50 response was evaluated at Week 24. The registry reports chi-squared analyses for the olokizumab-versus-placebo comparisons and a risk difference for the adalimumab-versus-placebo comparison.
| Comparison | Risk difference | CI | P-value | Hypothesis |
|---|---|---|---|---|
| Olokizumab q4w vs placebo | 0.275 | 97.5%: 0.192–0.349 | <0.0001 | Superiority |
| Olokizumab q2w vs placebo | 0.278 | 97.5%: 0.195–0.353 | <0.0001 | Superiority |
| Adalimumab q2w vs placebo | 0.237 | 95%: 0.165–0.303 | Not reported | Other / not stated |
The ACR50 endpoint requires a larger improvement than ACR20, so it provides a different view of response intensity. Statistically, however, the endpoint remains binary: each subject is classified according to whether the response criterion is met at Week 24.
11. Secondary Results: CDAI Remission at Week 24
The registry also reports the percentage of subjects achieving Clinical Disease Activity Index (CDAI) ≤ 2.8, classified as remission, at Week 24.
| Comparison | Risk difference | CI | P-value | Hypothesis |
|---|---|---|---|---|
| Olokizumab q4w vs placebo | 0.080 | 97.5%: 0.031–0.123 | 0.0003 | Superiority |
| Olokizumab q2w vs placebo | 0.069 | 97.5%: 0.020–0.111 | 0.001 | Superiority |
| Adalimumab q2w vs placebo | 0.089 | 95%: 0.046–0.127 | Not reported | Other / not stated |
The CDAI remission estimates are smaller in absolute magnitude than the Week 12 low-disease-activity estimates, illustrating why the definition of a binary endpoint matters. A treatment can produce a substantial difference in a broad disease-activity category while a more stringent remission endpoint shows a smaller risk difference.
For olokizumab q4w, the estimated risk difference was 0.080, with a 97.5% confidence interval of 0.031 to 0.123. For q2w, the estimate was 0.069, with a 97.5% confidence interval of 0.020 to 0.111.
Neither estimate should be converted into a relative effect without the underlying response proportions. The risk difference directly describes the absolute separation between groups on the probability scale.
12. Safety: Serious Adverse Events
The registry reports serious adverse events by randomized treatment arm as affected subjects divided by subjects at risk.
| Treatment arm | Serious adverse events | At risk |
|---|---|---|
| Olokizumab q4w + methotrexate | 20 | 477 |
| Olokizumab q2w + methotrexate | 22 | 463 |
| Adalimumab q2w + methotrexate | 26 | 462 |
| Placebo q2w + methotrexate | 12 | 243 |
These counts describe the number affected and the corresponding number at risk in each arm. They should not be interpreted as a formal between-group safety hypothesis test because the registry information reported here does not provide a comparative p-value, confidence interval, or prespecified multiplicity framework for these serious-adverse-event counts.
13. Statistical Methods Explained
Why was a chi-squared test used for ACR20?
ACR20 is binary at the analysis time point: a subject either meets the response definition or does not. A 2×2 chi-squared test compares the resulting response classification across two randomized treatment groups. The accompanying risk difference translates that categorical comparison into an absolute treatment-effect measure.
What does a risk difference of 0.270 mean?
An RD of 0.270 means the estimated probability of the specified outcome is 0.270 higher in the first-listed group than in the comparator. Expressed in percentage points, the difference is 27.0 percentage points. It is an absolute difference, not a 27.0% relative increase.
Why was ANCOVA used for HAQ-DI?
HAQ-DI is a continuous outcome, so the statistical problem differs from the binary ACR20 endpoint. ANCOVA provides a linear-model framework for comparing treatment groups while accounting for baseline information according to the prespecified model. The registry reports least squares mean difference as the effect measure.
Why is non-inferiority judged against a margin?
A non-inferiority trial asks whether the treatment is not unacceptably worse than the comparator by more than a prespecified amount. For CREDO 2, the registry gives a -12% margin for the ACR20 comparisons with adalimumab and a -7.5% margin for low disease activity. The confidence interval must therefore be interpreted relative to those margins rather than against zero alone.
Why does the ITT population matter?
Analyzing subjects according to randomized assignment maintains the comparison created by randomization. In CREDO 2, the registry explicitly identifies the ITT population as the primary analysis population and defines it as all randomized subjects analyzed according to their assigned treatment group.
What does a p-value of <0.0001 tell us?
It indicates that, under the null hypothesis and the reported statistical model or test, the observed result is associated with a very small tail probability. It does not tell us how large the treatment effect is, whether the effect is clinically important, or how precisely the effect has been estimated. Those questions require the effect estimate and its confidence interval.
14. Confidence Intervals and Effect Size
CREDO 2 illustrates why confidence intervals are essential alongside p-values. The primary q4w comparison has an estimated risk difference of 0.270 with a 97.5% confidence interval of 0.183 to 0.352. The q2w comparison has an estimate of 0.258 with a 97.5% interval of 0.171 to 0.341.
The visual comparison is descriptive rather than a formal test that the two olokizumab regimens differ from each other. Their confidence intervals overlap, and no direct q4w-versus-q2w hypothesis test is reported in the primary analyses listed here.
15. Multiplicity and the Two Primary Comparisons
The trial has one registered primary endpoint but two primary analyses: olokizumab 64 mg q4w versus placebo and olokizumab 64 mg q2w versus placebo. Both were superiority hypotheses evaluated with a 2×2 chi-squared framework.
| Primary analysis | Question | Hypothesis | Reported result |
|---|---|---|---|
| OKZ q4w vs placebo | Difference in Week 12 ACR20 response | Superiority | RD 0.270; 97.5% CI 0.183–0.352; P < 0.0001 |
| OKZ q2w vs placebo | Difference in Week 12 ACR20 response | Superiority | RD 0.258; 97.5% CI 0.171–0.341; P < 0.0001 |
When more than one confirmatory hypothesis is evaluated, the interpretation of type I error depends on the prespecified multiplicity strategy. The registry data identify both analyses as primary but do not provide a multiplicity-adjustment method in the statistical-analysis entries summarized here. The two reported p-values should therefore be understood as results of their respective registered analyses rather than used to infer an unreported error-control procedure.
16. Missing Data and the HAQ-DI Analysis
The handling of missing information is explicitly described for the HAQ-DI endpoint. Subjects with a missing baseline were not included. Data after treatment discontinuation were included, while data after discontinuing the study were multiply imputed using the return-to-baseline assumption.
Missing baseline
Subjects with a missing baseline measurement were not included in the HAQ-DI analysis.
After treatment discontinuation
Data after treatment discontinuation were included.
After study discontinuation
These data were multiply imputed based on the return-to-baseline assumption.
Statistical implication
The estimated least squares mean difference depends partly on these prespecified data-handling decisions.
Multiple imputation is not equivalent to simply replacing missing values with a single number. It is intended to incorporate uncertainty associated with missing observations. The return-to-baseline assumption is particularly important because it determines how observations after study discontinuation are represented in the analysis.
17. Non-Inferiority Analysis in Detail
The CREDO 2 registry entries provide an unusually clear illustration of the distinction between superiority and non-inferiority questions. For ACR20, the protocol-defined non-inferiority margin was 12%, represented by a lower confidence-limit threshold of -12%.
| Endpoint | Comparison | Estimate | Lower 97.5% CI | Margin |
|---|---|---|---|---|
| ACR20 | OKZ q4w vs adalimumab | 0.045 | -0.022 | -12% |
| ACR20 | OKZ q2w vs adalimumab | 0.034 | -0.035 | -12% |
| Low disease activity | OKZ q4w vs adalimumab | 0.076 | 0.004 | -7.5% |
| Low disease activity | OKZ q2w vs adalimumab | 0.069 | -0.003 | -7.5% |
For each comparison, the registry's stated criterion is based on whether the lower confidence limit exceeds the relevant negative margin. This is fundamentally different from asking whether the confidence interval excludes zero. A confidence interval can include zero and still satisfy a non-inferiority criterion if the entire interval remains above the unacceptable-loss boundary.
18. Trial Timeline
Trial start
The registry records the trial start date as June 6, 2016.
Randomized double-blind evaluation
The study evaluated two olokizumab regimens against placebo and adalimumab in a parallel randomized design.
Primary efficacy assessment
The registered primary endpoint was ACR20 response at Week 12, with superiority comparisons of each olokizumab regimen against placebo.
Later efficacy assessments
The registry reports ACR50 response and CDAI remission at Week 24.
Primary completion
The registry records August 2, 2019 as the primary completion date.
19. What the Primary Risk Difference Does — and Does Not — Mean
A risk difference of 0.270 means the estimated probability of meeting the ACR20 response definition was 0.270 higher in the olokizumab q4w group than in the placebo q2w group at Week 12.
Because the endpoint is binary, the effect is naturally expressed on the probability scale. It is not a hazard ratio, odds ratio, or relative risk.
The 97.5% confidence interval of 0.183 to 0.352 represents statistical uncertainty around the estimated risk difference. It does not describe the range of responses among individual subjects.
The p-value of <0.0001 describes evidence against the null hypothesis under the reported chi-squared analysis. It does not quantify the magnitude or clinical importance of the effect.
The ACR20 responder definition also requires the subject to remain on randomized treatment and in the study at Week 12. Consequently, the reported response difference reflects the registered composite endpoint rather than joint-count improvement alone.
20. Comparing Absolute and Model-Based Effects
CREDO 2 uses two particularly interpretable effect scales. Binary endpoints such as ACR20, low disease activity, ACR50, and CDAI remission are summarized with risk differences. The continuous HAQ-DI endpoint is summarized with a least squares mean difference.
| Endpoint type | Effect measure | Interpretive question |
|---|---|---|
| Binary response | Risk difference | How much higher or lower is the probability of meeting the endpoint? |
| Continuous HAQ-DI | Least squares mean difference | How different are the model-adjusted mean changes between groups? |
| Non-inferiority binary comparison | Risk difference + confidence interval | Does the confidence interval exclude a loss larger than the prespecified margin? |
This distinction matters because effect measures should match the structure of the endpoint. A p-value alone hides that distinction. The risk difference preserves the absolute scale of binary outcomes, while the least squares mean difference summarizes the adjusted continuous-outcome comparison.
21. Limitations and Interpretation Issues
- Binary endpoint construction: ACR20 requires both clinical response criteria and remaining on randomized treatment and in the study at Week 12. This should be considered when interpreting the response rate.
- Multiple comparisons: The registry reports two primary analyses of the same primary endpoint, one for each olokizumab regimen versus placebo. The statistical-analysis entries do not state a multiplicity-adjustment procedure.
- Non-inferiority margins: Non-inferiority conclusions depend on the prespecified margins and confidence intervals, not simply on whether a conventional p-value is statistically significant.
- Missing data: HAQ-DI results depend on the specified handling of missing baseline measurements and multiply imputed observations after study discontinuation.
- Unreported methods: Several secondary comparisons have a reported effect estimate and confidence interval but list the statistical method as not reported. The registry therefore does not provide a method-specific basis for describing those particular calculations beyond their reported effect measures and hypothesis types.
- Safety interpretation: Serious adverse-event counts are reported by arm, but the registry information summarized here does not provide comparative inferential statistics for those counts.
- Endpoint multiplicity: The trial includes several secondary efficacy endpoints at different time points. Their p-values should be interpreted in the context of the prespecified testing strategy; the registry entries summarized here do not provide a complete multiplicity hierarchy for every posted secondary analysis.
22. Why This Trial Matters Statistically
CREDO 2 is a useful statistical teaching example because it combines randomized superiority testing, active-control non-inferiority comparisons, binary response endpoints, a continuous patient-function endpoint, ITT analysis, and explicit missing-data handling.
| Concept | How it appears in CREDO 2 |
|---|---|
| Randomization | Four-arm randomized parallel-group design |
| Double masking | Double-blind trial design |
| ITT analysis | All randomized subjects analyzed according to randomized treatment group for primary efficacy |
| Chi-squared testing | Primary ACR20 and several secondary binary endpoint comparisons |
| Risk difference | Primary and secondary binary treatment-effect measure |
| ANCOVA | HAQ-DI comparison from baseline to Week 12 |
| Non-inferiority | Olokizumab-versus-adalimumab comparisons using prespecified margins |
| Confidence intervals | 97.5% intervals for primary and many secondary analyses |
| Missing-data methods | Multiple imputation using a return-to-baseline assumption for specified HAQ-DI observations |
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: CREDO 2, NCT02760407.
- PubMed: PMID 42142212.
- PubMed: PMID 41239458.
- PubMed: PMID 36001712.
Continue through the Clinical Biostats statistical library
Connect this trial's endpoint definitions and statistical methods to focused tutorials and practical statistical calculators.
26. Record Summary
CREDO 2 provides a compact example of several core clinical-trial statistical methods. The primary endpoint was a binary ACR20 response at Week 12, analyzed in the ITT population using chi-squared testing and risk difference. Both olokizumab regimens produced reported positive risk differences versus placebo, with 97.5% confidence intervals and p-values of <0.0001. The trial also incorporated active-control non-inferiority comparisons, where interpretation depends on the relationship between the lower confidence limit and a prespecified margin. Secondary analyses extended the statistical framework to low disease activity, ACR50 response, CDAI remission, and HAQ-DI, with ANCOVA used for the continuous HAQ-DI endpoint.
The statistical lesson is that the treatment question determines the effect measure and inferential framework. Superiority versus placebo is naturally described with the risk difference and its confidence interval; non-inferiority versus an active comparator requires comparison with a prespecified margin; and continuous outcomes require a model appropriate to their scale. Reading the estimate, confidence interval, p-value, analysis population, endpoint definition, and missing-data strategy together provides a more complete statistical interpretation than any one number alone.