Enhancing the Analysis of Designed Experiments Using Goodness-of-Prediction Crit...
Our World Statistics Day conversations have been a great reminder of how much statistics can inform our lives. Do you have an example of how statistics has made a difference in your life? Share your story with the Community!
Enhancing the Analysis of Designed Experiments Using Goodness-of-Prediction Criteria and Factor Importance Indices
Oct 9, 2017 7:32 AM
| Last Modified: Oct 26, 2017 10:13 AM
José G. Ramírez, PhD, Chief Statistician, Quality Excellence Group, Amgen
This paper was voted one of the three finalists for Best Contributed Paper.
After the analysis of a designed experiment with multiple responses, one may end up with a large number of reduced models that are a function of several experimental factors. How do we determine which of these models are good for prediction? How do we assess which experimental factors are the most important ones across all of the responses? The Box-Wetz criterion is a useful goodness-of-prediction measure for statistical models that compares the average change in a response to the average estimation error. The Sobol decomposition provides sensitivity indices that are very helpful in assessing the importance of the experimental factors across multiple responses. Using examples from the biotechnology industry, we will demo a JMP script that calculates the Box-Wetz criterion for a given model and discuss its interpretation. We will also show how the sensitivity indices generated by the “Prediction Profiler > Assess Variable Importance” can be used to prioritize the importance of experimental factors across multiple responses.