AI disclosure. This page was drafted by Grok Bot (AI) for Apiary from the sources listed below, which were opened and read. It has not been verified by the founder. Labels used: Observed result, Explanation, Editorial suggestion, Hypothetical. If something looks off, tell Austin — that's the point of a living hive.
Direct answer
A confidence interval (CI) is a range around an estimate, such as a population change, a colony-loss rate or a project effect. It shows how precisely the data pin that estimate down, if the statistical model's assumptions are correct. A narrow interval means a precise estimate. A wide one means the data are compatible with a broad range of values, possibly including "no change." A 95% CI does not mean there is a 95% chance the true value lies inside that particular interval. It also does not protect against bias from poor study design.
The evidence
What the 95% refers to (explanation, statistical guidance). Greenland and colleagues explain that if 95% intervals are calculated "repeatedly in valid applications, 95 % of them, on average, will contain ... the true effect size." That coverage "is a property of a long sequence of confidence intervals computed from valid models, rather than a property of any single confidence interval" (Greenland et al. 2016, Eur. J. Epidemiol. 31(4): 337–350; read). They list common misreadings:
- "The specific 95% interval has a 95% chance of containing the true value." No. Any one interval either contains it or doesn't. The 95% describes how often such intervals would, "if all the assumptions used to compute" them were correct.
- "Values outside the interval have been refuted." No. The interval "is computed from many assumptions, the violation of which may have led to the results."
- "If two intervals overlap, the difference isn't significant." No. Their example: intervals of (1.04, 4.96) and (4.16, 19.84) overlap, yet the test of no difference gives P = 0.03. Comparing groups "requires statistics that directly test and estimate the differences."
Values inside the interval are those "more compatible with the data" than values outside it, under the model.
Precise is not the same as accurate (simulation result, ecology). In Christie and colleagues' simulations on 47 ecological datasets, BACI estimates kept the true effect inside the 95% interval about 99% of the time. Simpler designs (before–after or control–impact only) lost coverage as sample size grew. Larger samples narrowed their intervals around a biased estimate: "increasing sample size tends to only increase precision in non-BACI designs around a biased estimate" (Christie et al. 2019, J. Appl. Ecol. 56(12): 2742–2754; read). A tight CI from a weak design can be confidently wrong.
CIs in practice (example, honey bee colony-loss surveys). The COLOSS standard methods for colony-loss surveys show how to report an overall loss proportion with a confidence interval. They use a quasi-binomial model to allow for extra variation between beekeepers (van der Zee et al. 2013, J. Apic. Res. 52(4); read). The interval reflects sampling uncertainty among respondents. It does not reflect who chose not to respond, which the same paper treats as a separate error source (non-response and selection bias).
How to read a CI in a report (editorial checklist)
- Find the estimate and the interval together. Example: "abundance change −12% (95% CI −30% to +6%)."
- Check whether it includes no effect (0% change, or a ratio of 1). If so, the data are compatible with no change and with sizable change, so don't report "no effect" as proven.
- Look at the width. Very wide intervals usually mean small samples or high variability. Plan more monitoring rather than drawing firm conclusions.
- Ask about design and assumptions. A CI covers random sampling error, not confounding, detection bias or non-response.
- Don't compare by eye. Overlapping intervals don't settle whether two groups differ. Look for an estimate of the difference itself with its own interval.
Limits
- Greenland et al. is a general statistics guide, written mainly for health research. The logic applies to ecological reports, but their examples are not ecological.
- Bayesian "credible intervals," common in some conservation models, have a different interpretation. They are not covered here.
Reader takeaway
A confidence interval tells you the range of values your data can reasonably support, given the analysis assumptions. Read it alongside the study design. Wide intervals mean uncertainty. Narrow intervals from a weak design can still be wrong.
Sources read for this answer
- Greenland S., Senn S.J., Rothman K.J., et al. (2016). Statistical tests, P values, confidence intervals, and power: a guide to misinterpretations. Eur. J. Epidemiol. 31(4): 337–350. https://doi.org/10.1007/s10654-016-0149-3
- Christie A.P., Amano T., Martin P.A., et al. (2019). J. Appl. Ecol. 56(12): 2742–2754. https://doi.org/10.1111/1365-2664.13499
- van der Zee R., Gray A., Holzmann C., et al. (2013). J. Apic. Res. 52(4). https://doi.org/10.3896/IBRA.1.52.4.18
AI disclosure: Drafted by Grok Bot, an AI assistant, for Apiary. No human expert review yet, and not founder-verified. Status: candidate.