Choose Language Hide Translation Bar
Level II

Using object properties in a model alongside multiple measurements of performance -- how best to model?

The situation is that a product can have multiple measurements of various properties, that may or may not be useful in predicting its performance in a separate test.  For each object, there are single measurements of properties, that truly describe the object, but multiple measurements of performance.  How can the properties best be included in a model of the performance?


Properties measured include attributes, which are numeric and continuous variables:

  1. Heat capacity
  2. Thermal conductivity
  3. Viscosity
  4. Traction
  5. Friction

Performance test is the number of stress cycles before failure.  Here, there are six replicates of this performance test.


So, for a given object, having singular property measurements, we have in this case six independent measures of performance.  So, experimental error is available, for the performance test, but not for the properties.  The experimental error gives a good estimate of variability to calculate the statistical significance.


It would seem quite possible that the properties are predictive of the performance, but it is impossible and impractical to make six measurements of these properties, just to match up to the number of performance tests.  The viscosity is, for example, truly known.  There is no reason to make six measurements.


How can JMP be used to link or employ the attribute properties in a model of performance?


I cannot create scripts.  Just saying.  Need JMP method independent of scripts...