In the random-effects model, the confidence interval of treatment variance/(treatment variance + random error variance) is relatively narrow because of the small number of looms used in the experiment.