Table 5:

Interpretation differences between NHST and Bayesian framework

	Probability p	Effect Interval [L, U]
NHST	If H0 is true, the probability of having the current result or more extreme is p (based on what would have occurred under other possible datasets); e.g., P(|T(y)| > tc|easy = difficult) = p, where T(y) is a statistic (e.g., Student’s t) based on data y and tc is a threshold.	If the study is exactly repeated an infinite number of times, the percentage of those confidence intervals will cover the true effectis 1 — p; e.g., P(L ≤ easy - difficult ≤ U) = 1 — p, where “easy - difficult” is treated as being fixed while L and U are random.
Bayesian	The probability of having the current result being different from zero is p (given the dataset); e.g., P(easy — difficult < L or easy — difficult > U|y) = p, where L and U are lower and upper bounds of the (1 — p)100% quantile interval.	The probability that the effect falls in the predictive interval is 1 — p (given the data); e.g., P(L ≤ easy — difficult ≤ U|y) = 1 — p, where “easy - difficult” is considered random while L and U are known conditional on data y.