Output from results used for the following studies:
- Trends and relevance in the bladder and bowel dysfunction literature: PlumX metrics contrasted with fragility indicators
- A second-look at reported statistics: Challenges in replicating reported p-values in pediatric urology literature
This code will calculate the 1) p-value fragility (for significant studies), 2) reverse p-value fragility (for non-significant studies), and 3) errors in reported p-value calculations.
To replicate the results:
- Install conda environment:
conda install --file conda.env
- Run the python scripts in order:
source pipeline.sh
The pipeline shell is as follows:
1_dataprep.py
: This script loads in the hard-coded excel files that have the list of studies with the different group counts and study covariates (study year, impact factor, study design). It will output~/output/{df_FI,df_inf}.csv
, the former having the group 1/2 counts, and the latter havng the study annotations.2_fragility.py
: Calculates the different FI index measures for all studies and saves results indf_res.csv
.3_results_FIFQ.py
: Calculates statistical associations on different journal factors and FI/FQ measures.4_results_insig.py
: Explores the papers that had baseline insignificant results and whether these could be accounted for.