nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

545048_en-US

liliconcom

Prediction Intervals for Mixed Models (Customer Request)

Hello everyone,

this is a follow up to Florians post a couple of time ago:

Solved: Prediction Interval in Prediction Profiler - JMP User Community

Here a work around is used to get the prediction interval for a model with fixed effects.

The question know is, if it is possible to calculate the prediction intervals for mixed models with random effects. Reading the post here it seems to be hard to get a JMP-formula for confidence intervals for mixed models because of the calculation method:

Get formula for "Mean Confidence Interval" from Fit Model for mixed effect model - JMP User Community

When I try to save formulas from the cholesterol stacked model in the sample library I am able to to safe the prediction formula and the confidence interval (without formula), but no PredSE for instance.

Is there a possible workaround to get the prediction interval formula or the confidence interval formula? Considering the post above, this seems to be not possible.

I am pretty new to mixed models, so I might have overlooked some obvious properties of mixed models, which make this hard/impossible. What would be the best answer to the customer?

Thanks,

Jonas

Re: Prediction Intervals for Mixed Models (Customer Request)

For mixed models you can get the prediction intervals (i.e., Indiv. Confidence Interval) when there are only random effects. For example, for cholesterol data, if you add Patient as a random effect and set the Repeated structure to Residual you can save the prediction interval values:

Mixed models with a Repeated structure other than Residual does not provide Indiv. Confidence Interval.

Since JMP does not provide the option to save Standard Error of Predicted as formulas, we cannot do the work-around as we did for SLS.

(BTW, JMP 17 now uses the t-quantile so that the prediction interval from the Prediction Profiler in Graph matches the saved Indiv Confidence Interval directly from Fit Model. I will update my previous post.)