A study has revealed a high prevalence of inconsistencies in reported statistical test results. Such inconsistencies make results unreliable, as they become “irreproducible”, and ultimately affect the level of trust in scientific reporting. statcheck is a free, open-source tool that automatically extracts reported statistical results from papers and recalculates p-values. Following an investigation into its accuracy, Michèle B. Nuijten finds […]
Category: p values
To understand the replication crisis, imagine a world in which everything was published.
Countering the claim that failed replications merely reflect the underlying truth of the scientific process and are not a matter of concern, Andrew Gelman argues that actually, the process is largely broken and we are in danger of dismissing the value of replication efforts. Current practice, centered on publication thresholds, is not filtering out poorly designed and executed studies. John Snow points me […]
Why the ban on P-Values? Understanding sampling error is key to improving the quality of research.
The weight placed on p-values and significance testing has come under increasing criticism, with one social psychology journal banning their use entirely. Nicole Radziwill argues that many of the issues come down to sampling errors. Inferential statistics is good because it lets us make decisions about a whole population based on one sample. But inferential statistics is bad if your sample size is too […]