Balancing Type I error and power in linear mixed models

This archive contains the simulation code (and results) for the article

Balancing Type I error and power in linear mixed models http://dx.doi.org/10.1016/j.jml.2017.01.001

Contents

1. sim.R - Contains basically all simulation and model selection code needed. mkdata() - Generates a data.frame containing every combination of condition, item and subject. simdat() - Samples one ore more responses given the variance/covariance parameters for every condition item and subject. fitModels() - Fits all models to the data. updateModels() - Re-fits all models to some new data (much faster than calling fitModels every time). simstats() - Assess model-statistics for a set of fitted models. simstep() - Performs a single simulation step. Samples new response, refits models and obtains model statistics. barr_modsel_backward() - Implements the LRT backward model selection. AIC etc. are rather trivial from the model statistics. Hence they are implemented in-line where needed.
2. pathological.R - Samples 10k responses from the minimal model are stores the results of all models in RDA files.
3. scan.R - Scans the SD of the random slopes in 20k steps and stores the results in RDA files.
4. modselPlot.R, pathologicalPrint.R and scanPlot.R - Output results.

Raw simulation results

The simulations take ages, hence we added our raw simulation results to this archiv (the RDA files). I you want to alter the simulations (e.g. altering the number of items/subjects or variances) you need to re-run the simulations by calling pathological.R or scan.R.