Sampling Bias in Research Data Collection

Sampling Design

Sample size n 30

Live Estimate

True population mean μ

—

Sample mean x̄

—

Error (x̄ − μ)

—

Repeated-Sample Reliability

Mean bias

run trials

RMSE

run trials

How it works

A spatial "research field" of N measurement units sits on the grid, each with a true value shaped by a smooth spatial trend plus a localized hotspot near one corner (visualized by height and color). The population mean is fixed and known here so every sampling design can be graded against it:

μ = (1/N) Σ v_i                (true population mean)
x̄ = (1/n) Σ v_i  over sample   (sample estimate)
Bias  = E[x̄] − μ               (systematic error over many samples)
RMSE  = √( E[(x̄ − μ)²] )       (typical size of the error, one draw)

Simple random — every unit has equal chance of selection; unbiased in expectation, but individual draws still scatter around μ.
Systematic — every k-th unit from a random start, in reading order; unbiased on this field, and usually lower variance than SRS because it spreads coverage evenly — unless the field has a periodic pattern that lines up with the interval.
Stratified — the field is split into four quadrants and sampled proportionally from each; removes the variance caused by the corner-to-corner trend, so RMSE over repeated trials is the lowest of the four.
Convenience — the n easiest-to-reach units nearest a fixed field station are collected instead of a probability sample; because the station sits near the hotspot, this design is biased — its repeated-sample mean bias stays far from zero no matter how large n gets, which more data cannot fix.

This is the core trade-off behind real research data collection: sample size controls variance (noise), but only the sampling design controls bias (systematic error) — a bigger convenience sample is still a biased one.

Population unit (height/color = value) Selected in current sample Field station (convenience anchor)

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.