Each dot is one volunteer's reported "colony health index" measurement, laid out on a horizontal value axis and jittered vertically only so overlapping dots stay readable (a beeswarm strip plot — identical math to the 3D version, viewed here as a flat distribution instead of a raised field of markers).
- Volunteer hives — more independent observations shrink random error through simple averaging, even if individual reports are noisy.
- Protocol rigor — a quick, low-effort protocol attracts more participants but produces noisier and more outlier-prone reports; a standardized, rigorous protocol reduces noise at a cost in volunteer time.
- Observer bias — untrained volunteers can share a systematic skew (consistently over- or under-reporting). Unlike random noise, bias does not average out as sample size grows — the aggregate line stays offset from the true-value line no matter how many volunteers report.
- QA filtering — flagging and excluding statistical outliers (more than 2 standard deviations from the raw mean) before analysis removes gross errors, at the cost of discarding some real, if unusual, data.
Well-run citizen science projects such as national beekeeping loss surveys can rival professional monitoring networks in statistical power simply because sample size is so large — but only if the protocol keeps bias in check, since averaging cannot fix a skew everyone makes in the same direction.