Date created: | Last Updated:

Category: Project

Description: A study assessing the degree to which effect sizes are inflated in the published literature looking at the large scale replication projects that have been performed. The easiest way to reproduce all analyses in this paper is to download and unzip "Reproducible_analysis_2019.08.16.zip" in the "analysis" component and run the .rmd file using R markdown. If you maintain the file structure from the zip file it should all run (assuming you have installed all required packages). The Bayesian mixture model is not run in this analysis script (because it takes ~ 20 minutes to complete on a fairly fast laptop), and the script for that can be found under the analysis folder as "mixtureModel.R". Don't hesitate to let me (Felix Singleton Thorn - fsingletonthorn@gmail.com) know if you see any bugs or if you need help re-performing any aspect of this analysis. Because the raw data for this analysis is from 8 separate studies, there are a number of different data files that are all in different formats. The final analysis data, which you are welcome to use for your own analyses of course, is called "estimating_the_effect_of_publication_bias_data.csv" and can be found in the data folder. See "estimating_the_effect_of_publication_bias_data_codebook.csv" for the codebook explaining each variable. The column labels should be self explanatory, but it will be helpful to know that the .r, .o and .ro suffixes on column names indicate that, respectively, the column contains data about the replication, original, or original-and-replication study.