Examining confidence intervals in Medline abstracts and full-texts.
The data Georgescu.Wren.RData has 1,320,226 rows and the following eight variables:
- "pubmed", pubmed ID number
- "journal", journal name
- "lower", lower confidence limit
- "mean", mean
- "upper", upper confidence limit
- "Year", year published
- "source", abstract or full-text
- "mistake", logical flag of whether it is a mistake because the mean is outside the limits
The full data file was too large to upload in text format. The file Georgescu.Wren.30.txt shows the first 30 rows of the data in tab-delimited format.
The Rmarkdown file 4_main_analysis_GW.Rmd performs all the analyses used in the paper, which are predominantly empirical cumulative density plots.
The R file 1_find.intervals.R is the code used to extract confidence intervals from pubmed abstracts written by me. This validates the more comprehensive code by Jonathan Wren, available here https://github.com/jdwren/ASEC.