Code to reproduce simulations in Lombardo, Lai, & Baron-Cohen (in press). Big data approaches to decomposing heterogeneity across the autism spectrum. Molecular Psychiatry. doi:10.1038/s41380-018-0321-0. https://www.nature.com/articles/s41380-018-0321-0

Simulations on effect size variability and inflation for simple case-control comparisons. Simulations will generate population-level data, and then over repeated simulated experiments, will randomly sample from those populations. Samples will be of different sizes. Simulations will also run across many population effect sizes. Finally, there is a simulation for calculating bias in sample prevalence of nested subtypes within one population. This is meant to illustrate how specific strata within the population can be enriched in studies that have small sample size.

Code implements the simulations in python or matlab.

effect_size_simulation.m or effect_size_simulation.py will do the main simulation to reproduce Figure 3A-E.

effect_size_inflation_sim.m reproduces Figure 3F.

heterogeneity_simulation.m or heterogeneity_simulation.py will do the simulation to reproduce Figure 4.