Is the Association Real, or Just Chance?
This reading contains every idea and every piece of evidence needed for today's decision. The research links at the end are optional.
Why this matters
Students can make better decisions when statistics stay connected to size, uncertainty, and study quality.
The question you are trying to answer
Which mark gives the best estimate?
Begin with the idea you already earned
Gene discovery combines broad scans, family structure, quality control, and independent replication.
Study the analogy before the biology
- Which mark gives the best estimate?
- What does the width of the band show?
- Why can a tiny gauge reading not prove the forecast is true?
Turn the analogy into three rules
Limit: Statistical uncertainty is not weather uncertainty, and its meaning depends on model assumptions.
Map those rules onto the biology
An effect size tells how large a difference or association is. A confidence interval shows a range of values compatible with the data and model.
A wide interval signals low precision. An interval that crosses the null value includes both no difference and effects in more than one direction.
A p-value asks how unusual the data would be under a stated null model. It does not measure importance, bias, or the chance that a claim is true.
Read Mateo's labeled case evidence
Study A reports a risk ratio of 0.60 with a 95 percent interval from 0.37 to 0.98.
The estimate suggests lower risk, with values near no difference still close to the interval edge.
Study B reports a risk ratio of 0.55 with an interval from 0.18 to 1.67.
The result is much less precise and includes possible benefit, no difference, or harm.
A p-value is calculated under a statistical model and null hypothesis.
It is not the probability that the scientific claim is true or false.
Make the concrete decision
You are the data reviewer preparing a one-slide result summary.
A teammate wants to label Study A true and Study B false based only on whether p is below 0.05.
- Compare effect sizes, intervals, designs, and bias before stating the claim.
- Use p below 0.05 as proof and ignore magnitude.
- Use the larger-looking effect as proof even when its interval is wide.
Choose the interpretation and explain why the two intervals support different precision claims.
Claim ceiling: You may describe magnitude, precision, and statistical compatibility. You may not convert a threshold into scientific truth.
Write the 10-year takeaway
Effect size and confidence interval show magnitude and precision; a p-value measures model compatibility, not truth.
- Which result is more precise, and why?
- What does a p-value not tell you?
Glossary in plain English
A statistical link where two things tend to occur together, which suggests a relationship but does not by itself prove one causes the other.
A number comparing the odds of having a factor in affected versus unaffected people; above 1 suggests the factor is associated.
A number showing how likely a result could happen just by chance; a smaller p-value makes luck a less believable explanation.
A range of values that likely contains the true result, showing how precise an estimate is; a narrow range means more certainty.
A result unlikely to be due to chance, often shown by a p-value below a set threshold such as 0.05.
A finding that shows no meaningful effect or difference, which is still valuable evidence and worth reporting.
Research citation trail (advanced)
You do not need these papers or database records to finish the lesson. They document where the plain-language explainer's claims come from and are intended for teachers or advanced readers.
- NEJM. Timing of Primary Surgery for Cleft Palate (TOPS), 2023.
- Gamble C, et al. 2023. Timing of Primary Surgery for Cleft Palate (TOPS trial). N Engl J Med. [PMID:37646677]
- Sabbagh HJ, et al. 2023. COVID-19 risk factors and orofacial clefts, five Arab countries: case-control study. BMC Oral Health. [PMID:37118740]
- Park JW, et al. 2007. Association between IRF6 and nonsyndromic cleft lip/palate in four populations. Genet Med. [PMID:17438386]