|
Abstract:
|
Power analyses are commonly needed to secure external funding from Federal, State, or local agencies. Many current software implementations for power analyses make assumptions in order for the calculations to be mathematically possible. Unfortunately, the data collected commonly do not meet the assumptions typically made within software used for power analyses. A more flexible and possibly more accurate power analysis framework can be achieved through Monte Carlo simulation. This presentation aims to introduce the benefits and flexibility achieved when conducting a power analysis under a Monte Carlo framework. The presentation will focus specifically on power analyses for general(-ized) linear (mixed) models using an R package called simglm (https://github.com/lebebr01/simglm). This package allows for flexible specification of data simulation conditions that may better approximate real data conditions. In addition, the package allows for the varying of data generating conditions, model fitting conditions, and others to explore impact of specific data characteristics on the empirical power.
|
