Relative Risk: Formula, Calculator, Interpretation, Python, R, SPSS and Excel Guide
Relative risk, also called the risk ratio, compares an event probability in one group with the event probability in a reference group. This complete guide explains the relative risk formula, how to calculate and interpret relative risk, confidence intervals, absolute versus relative risk, odds ratio comparisons, calculator checks, Python, R, SPSS, Stata and Excel workflows, APA reporting, ten unique chart interpretations and downloadable reports through a worked analysis of 649 students.
649 observations
RR = 0.7563
95% CI [0.6914, 0.8274]
Relative Risk Model Overview
Relative risk is the ratio of two event probabilities. It answers a proportional question: how large is the event risk in the numerator group compared with the event risk in the reference group? The term risk ratio is statistically equivalent and is often preferred because the estimate is a ratio rather than a vague measure of “relative” risk.
What Question Does It Answer?
The worked analysis asks whether the probability of G3 ≥ 10 differs between MS and GP students. MS is defined as the numerator or exposed group, GP is the reference group, and the event is passing the selected threshold.
How Relative Risk Is Built
First calculate the event proportion within each group. Then divide the MS risk by the GP risk. A value of 1 means equal risks, a value below 1 means the numerator group has a lower event risk, and a value above 1 means the numerator group has a higher event risk.
How Relative Risk Differs from an Odds Ratio
Relative risk compares probabilities, while an odds ratio compares odds. The distinction matters when the event is common. In this table, RR = 0.756 but the odds ratio is about 0.190, so substituting odds language for risk language would strongly exaggerate the proportional difference.
Worked Example at a Glance
Among 226 MS students, 158 reached G3 ≥ 10, giving a pass risk of 69.91%. Among 423 GP students, 391 reached the threshold, giving a pass risk of 92.43%. The MS-to-GP relative risk is 0.7563, with a 95% Katz log confidence interval from 0.6914 to 0.8274. The absolute MS-minus-GP risk difference is −22.52 percentage points.
Quick Answer: Relative Risk Result
The analysis compares the probability of G3 ≥ 10 in MS with the probability in GP. The risk ratio and its Katz log confidence interval are the primary results.
Estimate Summary
- Numerator group: MS
- Reference group: GP
- Event: G3 ≥ 10
- Relative risk: 0.7563316
- Interval method: Katz log confidence interval
Substantive Meaning
- MS: 158 of 226 passed.
- GP: 391 of 423 passed.
- MS risk was 24.37% lower on a relative scale.
- MS risk was 22.52 percentage points lower on an absolute scale.
- The confidence interval lies entirely below 1.
Table of Contents
- Research question, hypotheses and data design
- When to use relative risk
- Relative risk formula and calculation
- Variables and data dictionary
- Worked 2 × 2 table and complete results
- Relative risk versus odds ratio and related methods
- Relative risk calculator steps
- Ten unique Python and R chart interpretations
- Python, R, SPSS, Stata and Excel workflows
- Expandable software code
- Assumptions, limitations and advanced interpretation
- APA reporting for relative risk
- Common mistakes and corrections
- Reports and worked Excel download
- Related Salar Cafe guides
- Frequently asked questions
- Conclusion
Research Question, Hypotheses and Data Design
Research Question
How does the observed probability of G3 ≥ 10 in MS compare with the observed probability in GP?
Null and Alternative Hypotheses
| Hypothesis | Statistical statement | Meaning |
|---|---|---|
| Null hypothesis | H0: RR = 1 | The two group risks are equal. |
| Two-sided alternative | H1: RR ≠ 1 | The risks differ in either direction. |
| Directional alternative | H1: RR < 1 | The MS pass risk is lower than the GP risk; use only when prespecified. |
Sampling Design Behind the Estimate
Relative risk requires meaningful group denominators. Each student belongs to one school group, and the event proportion is calculated within that group. The table resembles a cohort comparison because both pass and non-pass counts are available for each school.
Why the Comparison Direction Must Be Declared
MS/GP = 0.7563, while GP/MS = 1.3222. These values are reciprocals and describe the same data from opposite directions. The numerator and reference groups must be specified before interpretation. See null and alternative hypothesis and Type I and Type II error.
When to Use Relative Risk
Use Relative Risk When
- The outcome is binary.
- Two independent groups are compared.
- Event probabilities can be estimated in both groups.
- Complete event and non-event denominators are available.
- The design is a cohort, randomized trial or suitable cross-sectional comparison.
- A probability ratio is more interpretable than an odds ratio.
- The comparison direction is prespecified.
- Absolute and relative effects can be reported together.
Choose Another Method When
- The study samples participants by outcome status, as in a typical case-control design.
- The same people are measured twice or matched.
- Observations are clustered and ordinary independent intervals are inadequate.
- The outcome has more than two categories.
- Time to event and censoring are central.
- Covariate adjustment is required.
- The group denominators do not represent risk sets.
- The exposure or outcome ordering is ambiguous.
Simple Test-Selection Logic
If no, ordinary relative risk may not be appropriate.
If no, use a paired, clustered or repeated-measures method.
If yes, consider log-binomial, modified Poisson or another regression model.
Relative Risk Is Not Restricted to Rare Outcomes
Relative risk remains directly interpretable whether an outcome is rare or common, provided risks are estimable. The rare-outcome condition is relevant when an odds ratio is used as an approximation to relative risk, not when RR itself is calculated.
Relative Risk and Outcome Dichotomization
The worked analysis converts G3 to pass versus non-pass. This answers a threshold question but discards differences within the two bands. A complete project can report the binary risk comparison and separately analyse the original quantitative outcome.
Relative Risk Formula and Calculation
The relative risk calculation has three core stages: calculate the event risk in each group, divide the numerator-group risk by the reference-group risk, and construct uncertainty on the logarithmic scale.
Step 1: Define the Four 2 × 2 Counts
Step 2: Calculate the MS Risk
Step 3: Calculate the GP Risk and Relative Risk
RR = 0.7563 means the MS pass risk is 75.63% of the GP pass risk. Equivalently, the observed MS pass risk is 24.37% lower on a relative scale.
Step 4: Calculate the Standard Error of ln(RR)
The Katz interval is calculated on the logarithmic scale using SE[ln(RR)] = √[(1/a − 1/(a+b)) + (1/c − 1/(c+d))] = 0.0458019.
Step 5: Calculate the 95% Confidence Interval
Substituting the observed values gives a 95% confidence interval of [0.6913940, 0.8273683].
Step 6: Calculate the Absolute Risk Difference
Step 7: Calculate the Relative Difference
The two effect scales complement each other: the observed MS pass risk is 22.52 percentage points lower and 24.37% lower relative to GP.
How to Interpret Common Relative Risk Values
| Relative risk | Basic interpretation | Required wording safeguard |
|---|---|---|
| RR = 1 | Equal observed risks | An interval may still show uncertainty around equality. |
| RR < 1 | Numerator-group risk is lower | State the numerator and reference groups. |
| RR > 1 | Numerator-group risk is higher | Do not convert association automatically into causation. |
| RR = 0 | No events in the numerator group when reference risk is positive | Ordinary log intervals may fail with zero cells. |
Variables and Data Dictionary
| Variable | Role | Coding | N | Counts | Meaning |
|---|---|---|---|---|---|
| school | Independent grouping variable | MS = numerator group; GP = reference group | 649 | MS = 226; GP = 423 | Defines the two risk denominators. |
| G3 | Original quantitative outcome | Final grade | 649 | Not analysed on its original scale here | Source used to create the pass threshold. |
| G3 ≥ 10 | Binary event | 1 = pass | 649 | 549 events | Event probability compared across schools. |
| G3 < 10 | Binary non-event | 0 = non-pass | 649 | 100 non-events | Completes each group denominator. |
Observed 2 × 2 Table
| School | G3 ≥ 10 | G3 < 10 | Total | Pass risk |
|---|---|---|---|---|
| MS | 158 | 68 | 226 | 0.6991150 |
| GP | 391 | 32 | 423 | 0.9243499 |
| Total | 549 | 100 | 649 | 0.8459168 |
2 × 2 Cell Labels Used in the Formula
| Cell | Count | Formula role | Interpretation |
|---|---|---|---|
| a | 158 | Numerator-group events | MS students with G3 ≥ 10. |
| b | 68 | Numerator-group non-events | MS students with G3 < 10. |
| c | 391 | Reference-group events | GP students with G3 ≥ 10. |
| d | 32 | Reference-group non-events | GP students with G3 < 10. |
Worked 2 × 2 Table and Complete Results
158 of 226
391 of 423
MS divided by GP
Katz log interval
Katz log interval
MS minus GP
Primary Relative Risk Result
| Measure | Estimate | 95% lower | 95% upper | Interpretation |
|---|---|---|---|---|
| Relative risk, MS/GP | 0.7563316 | 0.6913940 | 0.8273683 | MS has a lower observed pass risk. |
Effect and Comparison Estimates
| Measure | Estimate | 95% lower | 95% upper | Interpretation |
|---|---|---|---|---|
| MS pass risk | 0.6991150 | — | — | Observed probability of G3 ≥ 10 in MS. |
| GP pass risk | 0.9243499 | — | — | Observed probability of G3 ≥ 10 in GP. |
| Risk difference, MS − GP | −0.2252348 | −0.2901235 | −0.1603462 | MS is 22.52 percentage points lower. |
| Relative difference, 1 − RR | 0.2436684 | — | — | MS risk is 24.37% lower relatively. |
| Odds ratio, MS/GP | 0.1901610 | 0.1201664 | 0.3009261 | Odds measure; not interchangeable with RR. |
Statistical Interpretation
The 95% confidence interval for the MS-to-GP relative risk lies entirely below 1. Under the Katz log interval, the data support a lower MS pass risk than GP pass risk.
Practical Interpretation
The absolute difference is large in straightforward percentage-point terms: 69.91% versus 92.43%, a gap of 22.52 points. The relative statement, RR = 0.756, adds that the MS risk is about three quarters of the GP risk. Report both because effect size interpretation depends on scale.
Confidence-Interval Interpretation
The interval [0.691, 0.827] describes plausible relative-risk values under the selected method and assumptions. It does not incorporate possible clustering, confounding, threshold sensitivity or measurement error.
Relative Risk Versus Odds Ratio, Risk Difference and Related Methods
Ratio of two event probabilities; primary measure in this guide.
Ratio of two event odds; can look more extreme for common outcomes.
Absolute percentage-point gap between the two event probabilities.
Tests association but does not replace an interpretable effect estimate.
Numerical Comparison for the Worked Table
| Method or measure | Estimate/statistic | Interval or p-value | Interpretive role |
|---|---|---|---|
| Relative risk, MS/GP | 0.7563316 | 95% CI [0.6913940, 0.8273683] | Primary proportional probability comparison. |
| Risk difference, MS − GP | −0.2252348 | 95% CI [−0.2901235, −0.1603462] | Absolute percentage-point comparison. |
| Odds ratio, MS/GP | 0.1901610 | 95% CI [0.1201664, 0.3009261] | Odds comparison; much farther from 1. |
| Pearson chi-square | χ²(1) = 57.3314 | p < .001 | Global unadjusted association test. |
| Fisher exact test | OR = 0.1902 | p < .001 | Conditional exact association comparison. |
| Feature | Relative risk | Odds ratio | Risk difference | Association test |
|---|---|---|---|---|
| Scale | Ratio | Ratio of odds | Absolute difference | Test statistic and p-value |
| Null value | 1 | 1 | 0 | Depends on reference distribution |
| Direct probability meaning | Yes | No | Yes | No effect magnitude by itself |
| Useful for common outcomes | Yes | Requires careful interpretation | Yes | Yes for testing |
| Main limitation | Needs estimable risks and group denominators | Often misread as risk | Depends strongly on baseline risk | Can encourage p-value-only reporting |
Relative Risk Versus Odds Ratio
The search phrase relative risk vs odds ratio reflects a common interpretation problem. RR = 0.756 means the MS probability is 75.6% of the GP probability. OR = 0.190 means the MS odds are 19.0% of the GP odds. Because passing is common, the odds ratio is much farther from 1 and must not be described as a probability ratio.
Relative Risk Versus Absolute Risk
Absolute vs relative risk is not an either-or decision. The relative result is 24.37% lower, while the absolute result is 22.52 percentage points lower. Absolute effects often communicate practical magnitude more directly.
Should Every Method Be Reported?
No. Report relative risk and absolute risk difference as the main effect measures. A chi-square test or Fisher’s exact test can be included as supporting inference, but method shopping based on the smallest p-value should be avoided.
Relative Risk Calculator: Step-by-Step Workflow
A reliable relative risk calculator must identify the event, numerator group and reference group before dividing probabilities. It should also provide a confidence interval and retain the four underlying counts.
Calculator Inputs for the Worked Example
| Input | Value | Meaning |
|---|---|---|
| Numerator-group events | 158 | MS students with G3 ≥ 10. |
| Numerator-group non-events | 68 | MS students with G3 < 10. |
| Reference-group events | 391 | GP students with G3 ≥ 10. |
| Reference-group non-events | 32 | GP students with G3 < 10. |
| Confidence level | 95% | Uses z = 1.959964 for the Katz interval. |
| Comparison direction | MS/GP | Determines the reported ratio and wording. |
Calculator Step 1: Confirm the Group Totals
The calculator must return 226 for MS and 423 for GP. If it uses 549 and 100 as group totals, the outcome categories have been mistaken for the groups.
Calculator Step 2: Confirm the Two Risks
The expected group risks are 0.699115 for MS and 0.924350 for GP. A risk should never be calculated from the grand total when the research question concerns a within-group probability.
Calculator Step 3: Confirm the Relative Risk
MS divided by GP should equal 0.7563316. Reversing the group direction should return approximately 1.322171, the reciprocal.
Calculator Step 4: Inspect the Confidence Interval
The Katz log interval should be approximately [0.691394, 0.827368]. A calculator should identify the interval method and warn when zero cells make the ordinary log formula unstable.
Calculator Step 5: Interpret Absolute and Relative Effects
Read RR beside the −22.52-percentage-point risk difference. A calculator that returns only a ratio without the underlying risks is incomplete for applied reporting.
Ten Unique Python and R Relative Risk Chart Interpretations
The supplied files contain ten unique chart URLs. The observed-table figure is shared by Python and R and is displayed once, while every other Python-only and R-only chart is displayed and interpreted separately.
Chart 1: Relative Risk Observed 2 × 2 TablePython + R
MS contains 158 passes and 68 non-passes, while GP contains 391 passes and 32 non-passes.
MS: 158/226 pass. GP: 391/423 pass. Total N = 649.
The table supplies the four counts used in every risk, relative risk, confidence interval and comparison calculation.
Relative risk uses group-specific denominators. Comparing only 158 with 391 would confound risk with unequal group size.
Chart 2: Python Group Risk ComparisonPython
The GP pass-risk bar is substantially higher than the MS pass-risk bar.
MS risk: 0.6991 or 69.91%. GP risk: 0.9243 or 92.43%.
MS minus GP equals −0.2252, a difference of −22.52 percentage points.
The risks are calculated from each NumPy or pandas row total, not from the grand total of 649.
Chart 3: Python Relative Risk with 95% Confidence IntervalPython
RR = 0.7563 when MS is compared with GP.
The 95% Katz interval is [0.6914, 0.8274].
The entire interval lies below 1, supporting a lower observed pass risk in MS than GP.
The interval is relatively narrow because the table includes 549 passes and 100 non-passes across 649 observations.
Chart 4: Python Outcome Counts by GroupPython
There are 158 MS passes and 391 GP passes.
There are 68 MS non-passes and 32 GP non-passes.
GP has more passes despite also having the larger sample; the decisive comparison is the within-group pass proportion.
Displaying both event and non-event counts prevents a risk ratio from being detached from the underlying data.
Chart 5: Python Relative Risk Result SummaryPython
RR = 0.7563 with 95% CI [0.6914, 0.8274].
MS = 69.91%; GP = 92.43%.
MS minus GP = −22.52 percentage points.
The confidence interval excludes 1, so equal observed risks are not supported by the log-RR interval.
Chart 6: R School-Specific Pass RisksR
The R output reproduces MS risk = 0.6991 and GP risk = 0.9243.
Rows identify schools and columns identify pass versus non-pass outcomes.
MS has about 75.6% of the GP pass risk under the selected comparison direction.
Use explicit dimnames so the numerator, denominator and event category remain visible in the code and output.
Chart 7: R Relative Risk ComponentsR
MS risk = 158/(158 + 68) = 0.699115.
GP risk = 391/(391 + 32) = 0.924350.
ln(RR) = −0.279275 and SE[ln(RR)] = 0.045802.
A negative log relative risk corresponds to RR below 1 and a lower numerator-group risk.
Chart 8: R Relative Risk Confidence IntervalR
The R estimate is 0.7563316.
The lower confidence limit is 0.6913940.
The upper confidence limit is 0.8273683.
Because both limits are below 1, the interval indicates a lower MS pass risk under the stated comparison.
Chart 9: R Absolute and Relative Risk ComparisonR
The MS pass risk is 22.52 percentage points lower than the GP pass risk.
RR = 0.7563, corresponding to a 24.37% lower relative pass risk.
Absolute and relative summaries answer different questions and should be reported together.
The phrase “risk reduction” can imply intervention causality; in this observational school comparison, use descriptive wording.
Chart 10: R Relative Risk Result SummaryR
MS: 158 passes and 68 non-passes; GP: 391 passes and 32 non-passes.
Relative risk = 0.7563.
95% CI [0.6914, 0.8274].
The observed MS pass probability is lower, but the table alone does not establish why the school profiles differ.
Relative Risk in Python, R, SPSS, Stata and Excel
Relative Risk in Python
Python can calculate relative risk directly from a NumPy array or pandas table. SciPy supports comparison tests, while the RR and Katz interval remain transparent formula calculations.
- Keep rows as groups and columns as event/non-event.
- Calculate each row risk separately.
- Use the log scale for the interval.
- Report the four counts and group labels.
- Calculate absolute risk difference beside RR.
Relative Risk in R
R can calculate RR from a labelled matrix or use epidemiology packages. Explicit manual formulas make the direction and denominators easy to audit.
- Assign dimnames to rows and columns.
- Extract risks from the correct rows.
- Build the Katz interval on ln(RR).
- Retain risk difference and relative difference.
- Check package orientation before reporting.
Relative Risk in SPSS
SPSS CROSSTABS can produce row percentages, chi-square output and risk estimates for a 2 × 2 table.
- Verify row and column category ordering.
- Request RISK and CHISQ statistics.
- Read the displayed risk estimate label carefully.
- Do not assume the first ratio uses the desired event.
- Export the crosstab with counts and row percentages.
Relative Risk in Stata
Stata’s cohort-study command can calculate risk ratios and related measures once exposure and event variables are coded as binary.
- Code GP as the reference exposure category.
- Code pass as the event.
- Use
csfor cohort-style measures. - Use tabulation for visible row risks.
- Confirm that the reported direction is MS versus GP.
Relative Risk in Excel
Excel can reproduce every primary calculation with visible formulas and is useful for teaching, audit and calculator-style work.
- Enter the four counts in a clearly labelled table.
- Calculate group totals with SUM.
- Calculate each risk from its row denominator.
- Use LN and EXP for confidence limits.
- Add checks for zero event cells.
Expandable Code for Relative Risk
Python: relative risk and Katz confidence interval
import math
from scipy.stats import norm
a = 158 # MS pass
b = 68 # MS non-pass
c = 391 # GP pass
d = 32 # GP non-pass
risk_ms = a / (a + b)
risk_gp = c / (c + d)
relative_risk = risk_ms / risk_gp
se_log_rr = math.sqrt(
(1 / a) - (1 / (a + b)) +
(1 / c) - (1 / (c + d))
)
z_critical = norm.ppf(0.975)
lower = math.exp(
math.log(relative_risk) - z_critical * se_log_rr
)
upper = math.exp(
math.log(relative_risk) + z_critical * se_log_rr
)
print(risk_ms)
print(risk_gp)
print(relative_risk)
print(lower, upper)Python: absolute effects and comparison tests
import math
from scipy.stats import chi2_contingency, fisher_exact
table = [[158, 68],
[391, 32]]
risk_ms = 158 / 226
risk_gp = 391 / 423
risk_difference = risk_ms - risk_gp
relative_risk_reduction = 1 - (risk_ms / risk_gp)
odds_ratio = (158 * 32) / (68 * 391)
chi2, p_chi, df, expected = chi2_contingency(
table,
correction=False
)
fisher_or, p_fisher = fisher_exact(
table,
alternative="two-sided"
)
print(risk_difference)
print(relative_risk_reduction)
print(odds_ratio)
print(chi2, p_chi)
print(fisher_or, p_fisher)R: labelled 2 × 2 relative risk workflow
observed <- matrix(
c(158, 68,
391, 32),
nrow = 2,
byrow = TRUE,
dimnames = list(
school = c("MS", "GP"),
outcome = c("Pass", "Non-pass")
)
)
risk_ms <- observed["MS", "Pass"] / sum(observed["MS", ])
risk_gp <- observed["GP", "Pass"] / sum(observed["GP", ])
rr <- risk_ms / risk_gp
se_log_rr <- sqrt(
1 / observed["MS", "Pass"] -
1 / sum(observed["MS", ]) +
1 / observed["GP", "Pass"] -
1 / sum(observed["GP", ])
)
ci <- exp(log(rr) + c(-1, 1) * qnorm(.975) * se_log_rr)
risk_difference <- risk_ms - risk_gp
relative_risk_reduction <- 1 - rr
list(
risk_ms = risk_ms,
risk_gp = risk_gp,
relative_risk = rr,
confidence_interval = ci,
risk_difference = risk_difference,
relative_risk_reduction = relative_risk_reduction
)SPSS: crosstab and risk estimates
* Create a binary pass outcome: 1 = G3 >= 10; 0 = G3 < 10.
COMPUTE pass = (G3 >= 10).
VALUE LABELS pass 0 'Non-pass' 1 'Pass'.
* Ensure school categories are clearly labelled before analysis.
CROSSTABS
/TABLES = school BY pass
/FORMAT = AVALUE TABLES
/STATISTICS = CHISQ RISK
/CELLS = COUNT ROW
/COUNT ROUND CELL.
* Read the risk estimate only after checking the displayed
* row, column, event, and reference-category order.Stata: cohort-style risk ratio
generate byte pass = G3 >= 10
label define pass_lbl 0 "Non-pass" 1 "Pass"
label values pass pass_lbl
generate byte ms_exposure = school == "MS"
label define exposure_lbl 0 "GP reference" 1 "MS exposed"
label values ms_exposure exposure_lbl
cs pass ms_exposure
tabulate ms_exposure pass, row chi2 exactExcel: worksheet formulas
Suggested input cells
B3 = 158 MS pass
B4 = 68 MS non-pass
C3 = 391 GP pass
C4 = 32 GP non-pass
MS risk:
=B3/SUM(B3:B4)
GP risk:
=C3/SUM(C3:C4)
Relative risk:
=(B3/SUM(B3:B4))/(C3/SUM(C3:C4))
SE of ln(RR):
=SQRT(1/B3-1/SUM(B3:B4)+1/C3-1/SUM(C3:C4))
95% CI lower:
=EXP(LN(Relative_Risk_Cell)-NORM.S.INV(0.975)*SE_Log_RR_Cell)
95% CI upper:
=EXP(LN(Relative_Risk_Cell)+NORM.S.INV(0.975)*SE_Log_RR_Cell)
Risk difference, MS minus GP:
=B3/SUM(B3:B4)-C3/SUM(C3:C4)
Relative risk reduction:
=1-Relative_Risk_CellAssumptions, Limitations and Advanced Interpretation
Core Assumptions
Data Conditions
- Two independent groups
- Binary event definition
- Complete group denominators
- One record per observation
- Correct group and event coding
- Meaningful risk-set interpretation
Reporting Conditions
- State numerator and reference groups
- Report both group risks
- Name the confidence interval method
- Report an absolute effect
- Avoid odds/risk confusion
- Avoid unsupported causal language
Independent observations
Each student should contribute one record to one school row and one pass-outcome column. Repeated records, matched observations or unresolved duplicates violate the ordinary two-group risk calculation.
Binary and mutually exclusive outcome
The event must be clearly defined as G3 ≥ 10 and the non-event as G3 < 10. Every observation must belong to exactly one of those categories.
Group denominators must represent the risk sets
The denominators 226 and 423 must be the complete analysed MS and GP groups. Excluding non-events after group formation would distort the risks.
Relative risk needs an interpretable sampling design
Risk ratios are most natural in cohort, trial or cross-sectional settings where group-specific event proportions can be estimated. Case-control sampling generally does not estimate population risks directly.
Reference-group choice changes the numerical ratio
MS/GP = 0.7563, while GP/MS = 1.3222. Both encode the same contrast, but the wording and confidence interval must match the chosen direction.
A confidence interval is not an assumption check
An interval excluding 1 describes sampling uncertainty under the model; it does not verify independence, data quality, absence of confounding or correct event coding.
Common outcomes make odds ratios look more extreme
The pass outcome is common in both groups. The odds ratio is about 0.190, much farther from 1 than RR = 0.756, so odds must not be described as risks.
Zero cells require special handling
The ordinary Katz standard error contains reciprocal cell counts. A zero event count can make the log estimate or interval undefined, requiring a justified correction or exact/model-based procedure.
Clustering can narrow naive intervals
Students may share classrooms, teachers or local conditions. If clustering is material, a simple 2 × 2 interval can understate uncertainty and a multilevel or robust approach may be needed.
Confounding remains possible
School and pass outcome may both be related to prior attainment, study time, absences or other variables. Relative risk from this table is unadjusted.
Effect modification may hide behind one overall ratio
A single overall risk ratio can mask different ratios across gender, prior achievement or another stratum. Stratified analysis or an interaction model can examine heterogeneity.
Causal language requires a causal design
The observed school comparison is descriptive. Relative risk does not by itself show that changing school would change a student's pass probability.
APA Reporting for Relative Risk
APA-style relative risk reporting should make the event, group direction, four counts, both risks, RR, confidence interval, absolute difference and causal limitations visible. The estimate is not interpretable when the numerator and reference groups are omitted.
What to Include in an APA-Style Report
Method Information
- Two independent groups
- Binary event definition
- Numerator and reference groups
- Confidence level and method
- Software or calculation procedure
- Observational or experimental design
Results Information
- Four observed counts
- Both group risks
- Relative risk and interval
- Absolute risk difference
- Relative difference where useful
- Noncausal practical interpretation
APA-Style Methods Sentence
“Relative risk was calculated to compare the probability of G3 ≥ 10 in MS with the corresponding probability in GP. MS was treated as the numerator group and GP as the reference group. A 95% Katz log confidence interval was calculated for the risk ratio.”
Full APA-Style Results Example
“The G3 ≥ 10 outcome occurred for 158 of 226 MS students (69.91%) and 391 of 423 GP students (92.43%). The MS-to-GP relative risk was 0.756, 95% CI [0.691, 0.827]. Thus, the observed MS pass risk was 24.37% lower on a relative scale and 22.52 percentage points lower on an absolute scale. These unadjusted observational results do not establish a causal effect of school.”
Concise APA Results Sentence
“The probability of G3 ≥ 10 was lower in MS (69.91%) than GP (92.43%), RR = 0.756, 95% CI [0.691, 0.827], an absolute difference of −22.52 percentage points.”
Nonsignificant Relative Risk
The event occurred in a/n₁ of Group 1 and c/n₀ of Group 0.
The relative risk was RR, 95% CI [lower, upper].
Because the interval included 1, the analysis did not provide sufficient evidence of unequal risks; it did not prove exact equality.
Lower Numerator-Group Risk
The event risk was risk₁ in Group 1 and risk₀ in the reference group.
The relative risk was RR below 1, 95% CI [lower, upper].
The numerator group had a (1 − RR) × 100% lower observed event risk. The statement describes association unless the design supports causation.
Higher Numerator-Group Risk
The event occurred in Group 1 count/total and reference count/total.
The relative risk was RR above 1, 95% CI [lower, upper].
The numerator group had a (RR − 1) × 100% higher observed event risk.
Reusable APA Templates for Relative Risk
APA Comparison with an Odds Ratio
“Relative risk was selected as the primary effect because group-specific event probabilities were directly estimable. The odds ratio was reported only as a secondary comparison and was not interpreted as a probability ratio.”
APA-Style Table for the Main Results
| Group or statistic | Value | 95% CI |
|---|---|---|
| MS G3 ≥ 10, n/N (%) | 158/226 (69.91%) | — |
| GP G3 ≥ 10, n/N (%) | 391/423 (92.43%) | — |
| Relative risk, MS/GP | 0.756 | [0.691, 0.827] |
| Risk difference, MS − GP | −0.225 | [−0.290, −0.160] |
| Relative difference, 1 − RR | 0.244 | — |
| Odds ratio, MS/GP | 0.190 | [0.120, 0.301] |
Suggested table note: “RR compares MS with GP. Confidence limits for RR use the Katz log method. The risk-difference interval uses a large-sample Wald method. Estimates are unadjusted.”
APA Rounding Rules
- Report RR and confidence limits to two or three decimals.
- Report both group percentages with one or two decimals.
- Use consistent precision across interval limits.
- Do not report a ratio without naming its direction.
- Use p < .001 only when a supporting test is actually below .001.
- Do not report p = .000.
APA Interpretation Words to Use
| Avoid | Use instead | Reason |
|---|---|---|
| “MS caused fewer passes.” | “MS had a lower observed pass risk.” | The table is observational and unadjusted. |
| “Students were 75.6% less likely.” | “MS risk was 75.6% of GP risk, or 24.4% lower.” | RR itself is the retained proportion, not the percentage reduction. |
| “The odds ratio was the relative risk.” | “RR was 0.756; OR was 0.190.” | Odds and probability are different scales. |
| “The groups were different because the interval was narrow.” | “The interval excluded 1 and was relatively precise.” | Precision and effect magnitude are separate. |
| “Relative risk reduction proves benefit.” | “The observed numerator-group risk was lower.” | Benefit requires a meaningful intervention and causal design. |
Common Relative Risk Mistakes and How to Correct Them
| Common mistake | Why it is a problem | Correct approach |
|---|---|---|
| Dividing event counts directly | 158/391 ignores unequal group denominators. | Divide each event count by its own group total first. |
| Failing to name the reference group | The ratio and interpretation become ambiguous. | State MS/GP or GP/MS explicitly. |
| Calling RR = 0.756 a 75.6% reduction | The reduction is 1 − RR = 24.4%. | Say the risk is 75.6% as large or 24.4% lower. |
| Using odds and risk interchangeably | The outcome is common and OR = 0.190 differs sharply from RR. | Use probability language for RR and odds language for OR. |
| Ignoring absolute risk | A ratio alone can obscure the 22.52-point gap. | Report the risk difference beside RR. |
| Using causal “risk reduction” language automatically | The school comparison is observational. | Use descriptive association wording. |
| Applying the log formula with a zero cell | Reciprocal terms become undefined. | Use a justified correction or another interval method. |
| Using relative risk in an outcome-sampled case-control design | Population risks are not directly estimated. | Use an odds ratio or a design-appropriate model. |
| Ignoring clustering or confounding | The unadjusted interval may overstate precision or misattribute association. | Use robust, stratified or regression methods when needed. |
| Reporting only a confidence-interval decision | The reader cannot see counts, risks or practical magnitude. | Report the complete effect summary. |
Relative Risk Reports and Worked Excel Download
R Relative Risk ReportLabelled matrix, risk components, confidence interval and R figures.
SPSS Relative Risk OutputCrosstab counts, row percentages, association tests and risk estimates.
Worked Excel AnalysisFormula-audited counts, risks, relative risk, interval and reporting sheet.
Frequently Asked Questions About Relative Risk
The questions below address the relative risk formula, interpretation, calculator checks, confidence intervals, software, odds-ratio comparisons and reporting.
What is relative risk?
Relative risk, also called the risk ratio, is the event probability in one group divided by the event probability in a reference group.
Is relative risk the same as risk ratio?
Yes. In epidemiology and applied statistics, relative risk and risk ratio usually refer to the same quantity.
What is the relative risk formula?
RR = [a/(a+b)]/[c/(c+d)], where a and b form the numerator group and c and d form the reference group.
What is the relative risk in this example?
The MS-to-GP relative risk is 0.7563316.
How do I interpret RR = 0.756?
The observed MS pass risk is about 75.6% of the GP pass risk, or about 24.4% lower on a relative scale.
What are the group risks?
MS pass risk is 158/226 = 69.91%; GP pass risk is 391/423 = 92.43%.
What is the 95% confidence interval?
The Katz log-scale 95% confidence interval is [0.6914, 0.8274].
What is the null value for relative risk?
The null value is 1 because RR = 1 indicates equal risks in the two groups.
Does the interval exclude 1?
Yes. Both limits are below 1 under the MS-to-GP comparison.
What is the absolute risk difference?
MS minus GP is −0.2252, or −22.52 percentage points.
What is relative risk reduction here?
One minus RR equals 0.2437, or 24.37%; use descriptive wording because this is not a randomized intervention.
Can relative risk be greater than 1?
Yes. RR greater than 1 means the numerator group has a higher event risk than the reference group.
Can relative risk equal 0?
RR can be 0 when the numerator group has zero events and the reference risk is positive, although interval estimation then needs special handling.
Can relative risk be negative?
No. Risks are nonnegative, so a valid relative risk cannot be negative.
Why does group order matter?
Reversing the groups takes the reciprocal: MS/GP = 0.7563, while GP/MS is about 1.3222.
Is relative risk the same as odds ratio?
No. Relative risk compares probabilities; an odds ratio compares odds. They can differ greatly when the outcome is common.
What is the odds ratio in this table?
The MS-to-GP odds ratio is approximately 0.1902, which should not be described as a 0.1902 risk ratio.
When should I use relative risk?
Use it when group-specific event probabilities can be estimated from an appropriate cohort, trial or cross-sectional design.
Can I use relative risk in a case-control study?
Ordinary case-control sampling fixes outcome counts and generally does not estimate population risk directly, so the odds ratio is usually used.
Does relative risk prove causation?
No. Causal interpretation depends on design, exchangeability, measurement and control of confounding.
How do I calculate relative risk in Python?
Calculate each row risk from its own denominator, divide the numerator-group risk by the reference risk, and use the log scale for the Katz interval.
How do I calculate relative risk in R?
Use a labelled 2 × 2 matrix, calculate row risks, divide them in the intended direction and exponentiate the log-scale interval limits.
How do I calculate relative risk in SPSS?
Use CROSSTABS with RISK statistics, then verify which displayed category and group define the reported risk estimate.
Can Excel calculate relative risk?
Yes. Excel can calculate risks, RR, the log standard error and Katz confidence limits with transparent worksheet formulas.
What should an APA report include?
Report the four counts, both risks, comparison direction, RR, confidence interval, absolute difference and a noncausal interpretation.
How should RR = 0.7563316 be rounded?
RR = 0.756 or RR = 0.76 is usually sufficient in prose, with consistent precision for the confidence interval.
What if a cell count is zero?
The ordinary log formula may fail. Use a justified continuity correction, exact method or appropriate regression model.
What if observations are clustered?
Use a method that accounts for clustering, such as robust standard errors, generalized estimating equations or a multilevel model.
Should I report absolute and relative risk?
Yes. The percentage-point difference and relative ratio provide complementary practical and proportional interpretations.
What is the main conclusion?
MS students had a lower observed pass risk than GP students: RR = 0.756, 95% CI [0.691, 0.827], with an absolute difference of −22.52 percentage points.
Relative Risk Conclusion
Relative risk is a direct and interpretable measure for comparing event probabilities in two groups when both group denominators are available. It should always be presented with the event definition, numerator and reference groups, four observed counts and both underlying risks.
In the worked example, 158 of 226 MS students and 391 of 423 GP students reached G3 ≥ 10. The MS-to-GP relative risk was 0.756, 95% CI [0.691, 0.827]. The observed MS pass risk was therefore about 24.4% lower on a relative scale and 22.52 percentage points lower on an absolute scale.
The association is unadjusted and does not establish a causal school effect. A strong report combines the relative and absolute effects, distinguishes risk from odds, considers clustering and confounding, and uses regression or stratified methods when adjustment is required.
