P-value calculator

Many published findings with p < 0.05 fail to replicate. A landmark 2015 study found only 36% of psychology experiments replicated successfully. P-values don't guarantee truth — they're one piece of evidence in a larger process.

A 95% confidence interval tells you the range of plausible effect sizes, not just whether the effect is "significant." An effect of 5 ± 1 is more informative than "p = 0.03." Many statisticians recommend reporting both.

One-tailed: you predict a direction (A > B). Two-tailed: you allow for either direction (A ≠ B). One-tailed tests are more powerful but require a priori directional hypotheses. Using one-tailed post hoc is considered p-hacking.

Bayesian statistics calculate the probability that your hypothesis is true given the data — what most people think p-values represent. Bayesian methods require specifying prior beliefs but provide more intuitive conclusions.

If the null hypothesis is true, there's a 5% chance of observing data as extreme as yours. It does NOT mean there's a 5% chance the null is true. P-values are about data probability, not hypothesis probability.

Conventionally, p < 0.05. This means your data would be unlikely (<5% chance) if nothing interesting were happening. But the 0.05 threshold is arbitrary — practical significance matters more than statistical significance.

A large p-value (e.g., p = 0.40) means your data is consistent with the null hypothesis. It doesn't prove the null is true — absence of evidence isn't evidence of absence. You simply don't have enough evidence to reject it.

One-tailed tests check for an effect in one direction only (is A greater than B?). Two-tailed tests check for any difference (is A different from B?). Two-tailed is more conservative and usually appropriate unless you have a strong prior directional hypothesis.