NHST |
Tests if an effect is statistically significant (p-value based) |
Simple, widely used |
Over-reliance on p-values, arbitrary thresholds (e.g., 0.05) |
TOST |
Tests whether two means are sufficiently close (equivalence) |
Useful in equivalence testing |
Requires pre-specified equivalence margin |
Bayesian Testing |
Uses posterior probabilities and Bayes Factor |
Incorporates prior knowledge, intuitive |
Requires prior distribution, computationally expensive |
Decision-Theoretic |
Minimizes expected loss based on cost functions |
Practical for decision-making |
Needs subjective cost assignment |
False Discovery Rate |
Controls proportion of false positives in multiple tests |
Useful in high-dimensional data |
Can still lead to false discoveries |