bayesplain.core.normal_t.mean_posterior

bayesplain.core.normal_t.mean_posterior(n, mean, sd)[source]

Exact Student-t posterior for one unknown mean.

Parameters:
n : int

Number of observations.

mean : float

Sample mean.

sd : float

Sample standard deviation (n - 1 denominator).

Returns:

Frozen scaled Student-t with n - 1 degrees of freedom, centred on the sample mean with scale sd / sqrt(n).

Return type:

scipy.stats.rv_continuous_frozen

Notes

Under the reference prior this is exact. Its equal-tailed interval coincides numerically with the usual t confidence interval, which makes this the cleanest place to show that the two frameworks can agree on every digit and still license different sentences.