Update sample size and power analysis platform in DOE to handle sigmaCurrently the power analysis platform calculates the mean of the groups but often an engineer only knows sigma.
It would be much more intuitive to allow for sigma calculation from power analysis.
@sgilyard please feel free to clarify and expand as necessary.
IMFT JMP Day feedback session November 8, 2017
Re: Update sample size and power analysis platform in DOE to handle sigmaHi Stan. What I was referring to in the request was in reference to the k-means option in the sample size and power platform. Currently the k-means option requires the user to enter prospective means for the groups under study. It has been my experience(at least in my industry) that very often engineers do not prior data nor sufficient knowledge in many cases to generate an estimate of the prospective group means. What I have done in the past is a work around. Since the final power computation uses a delta computed by taking the square root of the sum of squares of each group mean minus the grand mean, you really only need to make the the group means have the right relative proportions to achieve the delta you want. In the past I've set-up a little formula in Excel to generate a set of group means using the equation for delta that would produce the delta I want. It would be easier if there was simply an option for a user to enter this delta rather than the group means. I suspect that the reason JMP used group means instead of the delta is because people might misunderstand the delta. When there's only two groups the delta is easy to interpret, since it's just the difference in the means between the two groups. When there are more than two groups, the delta is trickier to interpret because it an "overall" difference in the group means rather than the difference between any two group means. I think this could be handled by having a note or something for the user to view that describes what delta is and how it's computed.
Lately what I've been doing is simply using the Design Evaluation option in the Design Diagnostics platform instead. I just create a column with generic group labels and assign a certain number of rows to each group and then evaluate the design for power. This works but it's not ideal because you have to set-up a mock column and try different numbers of rows in each group until you get the power you want.
Let me know if you need any other info. Thanks.
Re: Update sample size and power analysis platform in DOE to handle sigmaHi Stan. Yes, I have comments on this topic. I remember what I was referring to regarding the Sample Size and Power platform. Let me put together what my suggestion/comments are and I will post by week's end. Thanks.
Re: Update sample size and power analysis platform in DOE to handle sigmaHi @sgilyard ,
In JMP 16 the Sample Size and Power platform is being overhauled.
Would you be able to provide some additional details here so that we might have an opportunity to work on the above request for JMP 16.
Looping in @calking to help with any additional questions from the JMP side.
Thanks,
Stan
Re: Update sample size and power analysis platform in DOE to handle sigma - Status changed to: In DevelopmentI believe that JMP 16 is taking an approach that will address this.
