Determining Confidence Limits for Linear Combinations of Variance Components in Mixed Models (2020-EU-EPO-455)
Jan 30, 2020 7:59 AM
| Last Modified: Apr 9, 2020 7:00 AM
Hadley Myers, JMP Systems Engineer, SAS Chris Gotwalt, JMP Director of Statistical Research and Development, SAS
Generating linear models that include random components is essential across many industries, but particularly in the Pharmaceutical and Life Science domains. The Mixed Model platform in JMP Pro allows such models to be defined and evaluated, yielding the contributions to the total variance of the individual model components, as well as their respective confidence intervals. Calculating linear combinations of these variance components is straightforward, but the practicalities of the problem (unequal Degrees of Freedom, non-normal distributions, etc.) prevent the corresponding confidence intervals of these linear combinations from being determined as easily. Previously, JMP Pro users have needed to turn to other analytic software solutions, such as the “Variance Component Analysis” package in R, to resolve this gap in functionality and fulfill this requirement. This presentation is to report on the creation of an add-in, available for use with JMP Pro, that uses parametric bootstrapping to obtain the needed confidence limits. The add-in, Determining Confidence Limits for Linear Combinations of Variance Components in Mixed Models , will be demonstrated, along with the accompanying details of how the technique was used to overcome the difficulties of this problem, as well as the benefit to users for which these calculations are a necessity.