Fig. 4

Main simulation results (continuous outcome, single covariate, no interaction) for sample size 100. The number of repetitions for this simulation is 5000. The performance of analytic methods in terms of bias, coverage and power for the five different covariate–outcome relationships is displayed. The effect of treatment is 40. Model-based standard errors are indicated in black and empirical standard errors are shown in red. Estimates are indicated with \(\pm 1.96 \times\) Monte Carlo standard error bars. Note that the error bars are too small to be seen for power, due to the scale of the plots