Minitab Power and Sample Size: selecting a non-Wald interval method for one-sample proportion tests

Short answer

There is no dialog-level option in the Stat > Power and Sample Size > One Proportion interface to switch the interval method away from the default normal-approximation (Wald) basis. The only in-product route is to capture the Session Command Language (SCL) the dialog generates, edit the command to request a different method, and re-run it — but whether a Wilson or exact method is accepted, and whether it affects every output table, is version-dependent and must be verified on your installed release before you trust the numbers.

What is confirmed versus likely

Confirmed by inspection: the dialog itself exposes no interval-method control, and the SCL it emits can be viewed and edited. Likely, but version-sensitive: the underlying Power command accepts a method-related subcommand in some releases, while other releases silently ignore unrecognized specifications and still report Wald-based intervals. Treat the SCL route as conditional until you have verified it on your version (see below).

Recommended workflow

1. Run Stat > Power and Sample Size > One Proportion once with your real inputs (hypothesized proportion, comparison value, power, alpha).
2. Open the Session window / command line editor to see the SCL the dialog generated (in current desktop versions, enable the command line via the Session window, or use the session command history). Copy that command into the Command Line Editor.
3. Check the Minitab Help for your installed release (Help > search for the Power command's session command documentation, or the release notes) for a METHOD or interval-method subcommand on the one-proportion Power command. Only use subcommand names documented for your version — do not guess syntax, because unsupported keywords may be ignored without an error.
4. Re-run the edited command and compare the session output against the original dialog run. If the reported sample size or power is byte-for-byte identical, the method change was almost certainly ignored.
5. Sanity-check the result independently: for a proportion near 0 with small n, an exact-method sample size should be larger than the Wald-based one. If it is smaller or unchanged, do not rely on the output.

If SCL does not cooperate

Staying entirely inside Minitab is still possible without the Power dialog: use Stat > Basic Statistics > 1 Proportion, which in recent desktop versions does let you choose the exact (Clopper–Pearson) method for the confidence interval, and iterate candidate sample sizes manually against your power target. This is slower but uses a fully documented, dialog-supported path. A short external check (e.g., an exact binomial power calculation) is also a reasonable cross-validation even if the deliverable stays in Minitab.

Version caveats to verify before relying on output

• Whether the one-proportion Power command documents a method subcommand at all in your release.
• Whether the interval method propagates to all output — some reports and graphs may remain Wald-based even when the command-level method changes.
• Whether the change applies symmetrically to both sample-size and power calculations; verify each direction separately with a known test case.

One detail that would sharpen this answer: your exact Minitab version (e.g., 21.x vs. the current desktop release), since the supported subcommands and output behavior differ across releases. Without it, the safe recommendation is the verify-then-trust workflow above rather than any specific syntax.