Fig 1.
Power analysis by simulation an explainer.
Fig 2.
An explainer on factorial design and analysis.
Fig 3.
Results of simulations to explore impact of a baseline sex differences on statistical power.
(A) Illustrative plots of constructed datasets, ranging from zero to maximum baseline sex differences and treatment effect. (B) Statistical power from both a factorial and pooled analysis to detect a main effect of treatment when the main effect of sex varies (none = 0, small = 0.5, large = 1). (C) Power for each model term within factorial analysis output. (D) Power for post hoc comparison of control vs. treated within each sex independently. Simulation N = 1,000 for each scenario assessed. For the graphs of power (B–D), the horizontal dashed line indicates the target power. The data underlying this figure can be found in https://doi.org/10.5281/zenodo.7806724.
Fig 4.
The impact of a sex-dependent treatment effect on statistical power.
Results of simulations to calculate statistical power where there is an interaction between treatment and sex in the data where the interaction effect is in the same direction as the main treatment effect. (A) Illustrative plots of constructed datasets, ranging from no effect to maximum treatment by sex interaction (none = 0, small = 0.5, large = 1) for varying sizes of a main treatment effect (0–1). (B) Statistical power to detect treatment effect per size of interaction effect and statistical method. (C) Power for each model term within factorial analysis output. (D) Power for post hoc comparison control vs. treated within each sex. Simulation N = 1,000 for each scenario assessed. For the graphs of power (B–D), the horizontal dashed line indicates the target power. The data underlying this figure can be found in https://doi.org/10.5281/zenodo.7806724.
Fig 5.
The impact of a treatment effect specific to 1 sex on statistical power.
Results of simulations to calculate statistical power where there is an interaction driven by a treatment effect in 1 sex only. (A) Illustrative plots of constructed datasets, ranging from a zero to maximum treatment by sex interaction effect (0–2). (B) Statistical power to detect the main effect of treatment effect as a function of the interaction effect size for each statistical method. (C) Power for each model term within factorial analysis output. (D) Power for post hoc comparison control vs. treated within each sex. (E) Power for contrasting design strategies—10 males vs. 5 females and 5 males. Simulation N = 1,000 for each scenario assessed. For the graphs of power (B–E), the horizontal dashed line indicates the target power. The data underlying this figure can be found in https://doi.org/10.5281/zenodo.7806724.
Fig 6.
The impact of an opposite treatment effect on statistical power.
Results of simulations to calculate statistical power where there is an interaction driven by opposite treatment effect in each sex. (A) Illustrative plots of constructed datasets, ranging from no effect to maximum opposite effect in each sex (0–1 male, 0 to −1 female). (B) Statistical power to detect the main effect of treatment as a function of the interaction effect size for each statistical method. (C) Power for each model term within factorial analysis output. (D) Power for post hoc comparison control vs. treated within each sex. Simulation N = 1,000 for each scenario assessed. For the graphs of power (B–D), the horizontal dashed line indicates the target power. The data underlying this figure can be found in https://doi.org/10.5281/zenodo.7806724.
Table 1.
Details of how the simulation datasets were constructed to represent various biological situations.