Skip to content

Reconciling a mean interval with a two-sided test

A two-sided normal test and its matching interval lead to the same threshold conclusion under consistent definitions.

On this page

Reuse the known-sigma example

Use mean 50.4, known sigma 2 and n = 100 independent normal observations. The standard error is 0.2. For this exercise, use a 95% two-sided normal critical value of approximately 1.959964.

Calculate the bounds

The margin is 1.959964 × 0.2 = 0.3919928. Subtracting and adding gives approximately [50.0080072, 50.7919928]. The null value 50 lies below the lower bound.

Equivalent views
Interval calculationTest calculation
|50.4 − 50| > 1.959964 × 0.2|z| > 1.959964
0.4 > 0.39199282 > 1.959964

Match all settings before reconciling

The correspondence uses the same reference model, standard error, sidedness and significance level. A one-sided test and a 95% two-sided interval are not the same configuration. Mixing a normal interval with an unknown-sigma t test can also create a discrepancy.

Interpret the interval as a procedure

The confidence level describes repeated-sampling coverage of the method under its assumptions. This realized interval is a set of parameter values that satisfy its construction. It does not put a 95% posterior probability on the fixed mean without a separate probability model for that mean.

Further references