Significance Testing (P-Values)
Fisher (1925) and Neyman and Pearson (1933) give science its yes-or-no rule for evidence, the p-value and the 5% threshold, now used in most fields and blamed for many false findings.
Open in the interactive tree →Karl Pearson's chi-squared test (1900) produced the first formal p-value; Fisher's Statistical Methods for Research Workers (1925) made p below 0.05, one chance in twenty, the working limit for calling a result significant. Neyman and Pearson added alternative hypotheses and the two kinds of error (1928-33), the two camps quarrelled until Fisher's death in 1962, and textbooks merged their ideas from about 1940. In the early 1990s over 90% of articles in the Journal of Applied Psychology relied on such tests; in 2016 the American Statistical Association warned that a p-value is not the probability that a hypothesis is true.
Prerequisites
- Statistics & Least Squares1809Builds on the normal curve and the theory of errors.
- Field Trials and Randomization1925Fisher developed the tests to judge his Rothamsted experiments.
Unlocks
- Randomized Controlled Trial1948Trials decide whether a treatment works by a significance test.
- Genome-Wide Association Studies2005Genome-wide scans flag variants by p-value thresholds across millions of tests.
- Replication CrisisopenChance results that pass the p<0.05 filter are a core cause of failed replications.