topic Re: Power Details window option under Oneway platform, have others used this to illustrate power and sample size? in Discussions

Power Details window option under Oneway platform, have others used this to illustrate power and sample size?

Vball247 — Thu, 24 Sep 2026 19:32:05 GMT

Wondering if other JMP users have used the "Power Details window" under Oneway platform?

Following the talk "Can We Talk...About Sample Size?" (2026-US-30MP-2930) in Discovery Summit 2026 by Dale Human, thought it was very interesting how he said to check Power and suggested sample size, after, and before, an analysis. Dale suggested using the DOE > Sample Size Explorers > Power > Power for Two Independent Sample Means to check power.

Did not know that the Oneway platform has the option for "Power", and when you select it from the menu, shows a "Power Details window"! This is another "OMJ!" moment ("Oh My JMP already had it!")

Using the Fitness dataset, plotting Oxy vs Sex, performing two sample t-test, p-value is low, at 0.0057. But now check the Power of this test.

From the menu, if you select "Power", adds the "Power Details window" section to the Oneway window:

If you select "Solve for Power", it will give you the Power of this test.

Also, you can do a "wait-if" on increasing the number of samples, by steps. Looks like this:

And there is a small power plot at the bottom!

Recreating in the DOE > Sample Size Explorers > Power > Power for Two Independent Sample Means, get the same 0.8218 for Power.

Also interesting the script to enable this:

Oneway(
	Y( :Oxy ),
	X( :Sex ),
	Means( 1 ),
	Mean Diamonds( 1 ),
	Power(
		Alpha( 0.05 ),
		Sigma( 4.74003159066053 ),
		Delta( 2.53881137990873 ),
		Number( 31 ),
		Solve for Power,
		Done
	)
);

or for checking increments of sample size:

Oneway(
	Y( :Oxy ),
	X( :Sex ),
	Power(
		Alpha( 0.05 ),
		Sigma( 4.74003159066053 ),
		Delta( 2.53881137990873 ),
		Number( 31, 60, 5 ),
		Solve for Power,
		Done
	)
);

As anyone had experience using the other options?

- Solve for Least Significant Number

- Solve for Least Significant Value

- Adjusted Power and Confidence Level

I am looking to add this to our statistics training, wondered if anyone else has used this menu option in the Oneway platform.

Re: Power Details window option under Oneway platform, have others used this to illustrate power and sample size?

MRB3855 — Fri, 25 Sep 2026 12:51:34 GMT

Hi @Vball247 : Interesting stuff indeed. FWIW, a word of caution though about post hoc power calulations:

https://community.jmp.com/t5/Discussions/How-do-you-calculate-power-in-a-retrospective-cohort-study/m-p/709092#M89307

Re: Power Details window option under Oneway platform, have others used this to illustrate power and sample size?

Vball247 — Wed, 30 Sep 2026 14:17:41 GMT

Thanks @MRB3855. Good to know that power and p-value are directly related. I am not a statistician by training, just a long time process engineer.

I do propose teaching this power calculation in the context of volume manufacturing (Six Sigma training). I describe how we are never 100% sure of anything, and that more data is always better. I illustrate that p-value cut-off of 0.05 (5%) is somewhat arbitrary, we are humans, based on number 10, we like the number of 5, and it is convenient. JMP helps illustrate this with color: for 0.01 to 0.05 it is one color, for < 0.01 it is another color, just to point out the 0.01 to 0.05 range can still fool you.

Also beta is often 10, willing to take more risk with beta (missing a signal) than alpha (risk of a false signal and doing work we did not need to do), because in volume manufacturing, we never stop measuring. This is the principle of the SPC chart.

In context, for an SPC chart, we have the risk of reacting to a signal that is a false signal (alpha), even though in the SPC chart, the alpha is much lower. On the flip side, a manager will often ask that after work on a tool, to make sure you are back on target with the minimum number of runs. Some may say within 1 or 2 runs! I illustrate with simulation how even if a tool has a shift after the work, it takes time, based on the background noise, to pick up a signal that is truly there. This is where I like to illustrate the power. There may be a power of 0.90 to pick up a signal within 12 runs, based on the delta to detect and the background noise (standard deviation) in the Power/Sample size curves. But that is not a guarantee! The SPC chart may pick up sooner (you got lucky) or may miss it a little longer (you were unlucky), but more data will eventually pick it up (the sample size discussion).

That is why I agree that the power curve vs sample size curve is what is important, and not the post-hoc power number for your particular test.

I hope to follow up to this post with how the discussion goes in the next classes.