Feed supplement and beef from cows
Use this diagram to show that grouping experimental units into pairs that are similar (matched pairs) and randomly allocating the two treatments within each pair increases the accuracy of the estimate of the treatment effect.
Explain that it is usually impossible to avoid large differences between experimental units.
Firstly click Accumulate then repeat a completely randomised experiment several times, noting the accuracy of the estimate of the effect of the supplement.
Now select Paired by calf weight to group the calves into pairs by their weight at the start of the experiment. Repeat the paired experiment several times and note that the effect of the supplement is more accurately estimated.
The estimate from the paired experiment is as accurate as that on the previous page when all calves were similar weights.