Fig. 3.
Simulation results. Power (for a type I error rate of 1 × 10−3) for the LG test under five different true scenarios (left to right) is shown in the top row; the bottom row shows the proportion of times each model is selected as the best-fit model based on the BIC metric, given the replicate was significant for LG at p < 1 × 10−3. Black circles = basic; blue crosses = subset; brown diamonds = inv-subset; red plusses = modifier; green stars = general. See online supplementary figure 1 for full simulation results.