🔬 Beekeeping Research Methodology & Experimental Design Lab

About this simulation

Imagine testing whether a new supplemental feed increases honey yield. You place some hives on "treatment" feed and others as a control, wait through the season, and weigh the honey. But the field itself is never uniform — one edge might sit closer to a hedgerow with better forage, catch more sun, or sit in a wind-sheltered dip. If your treatment hives happen to cluster in the better spot, you cannot tell whether a yield difference came from the feed or from the location. This simulation lets you build that field trial yourself and watch how design choices change what you can honestly conclude.

How it works

• Blue hives are the control group; amber hives receive the treatment. Hive height and the floor's colour gradient encode each colony's simulated honey yield and the field's underlying forage quality.
• Randomised assignment scatters control and treatment hives evenly across the quality gradient, so the confound cancels out on average.
• Clustered assignment places every control hive on one side of the field and every treatment hive on the other — mimicking the common mistake of assigning by convenience (e.g. "new hives go on the empty side").
• The floating bar chart shows each group's mean yield ± standard error, recomputed live from simulated measurements.

Controls

• Assignment method — toggle between randomised and clustered (confounded) treatment allocation.
• Replicates per group — sample size per arm; more replicates shrink the standard error and make real effects easier to detect.
• True treatment effect — the actual honey-yield boost the feed produces, in kg. This is the number a good trial should recover.
• Site confound strength — how much the field's forage gradient alone changes yield, independent of any treatment.
• Colony-to-colony noise — natural random variation between colonies that no design can remove, only average out.
• ↺ Run new trial — redraws random colony outcomes with the current settings. Drag to rotate, scroll to zoom.

Did you know?

A rule of thumb in field ecology and apiculture research is that randomisation protects against confounds you didn't think to measure, while replication protects against confounds you can't remove — no amount of replication fixes a systematically biased design, and no amount of randomisation substitutes for enough colonies to see past natural variability.