Which Model When?
andreacoombs1
| Updated on
The models you use depend on your data, the questions you are trying to answer and the problems you want to solve. See how to decide by working through case studies that illustrate how to identify, fit and evaluate models that might be most useful in achieving your analysis goal.
Questions answered by Olivia Lippincott @O_Lippincott and Andrea Coombs @andreacoombs1 at the live webinar:
Q: Is the C-code a JMP-Pro feature, or is in available in JMP too?
A: The C-code is available in JMP Pro. The formula depot is a JMP Pro feature.
Q: How does the variable selection capability differ between the tree-based models and stepwise regression and LASSO? That is, how do we decide which of these models to use for feature reduction purposes?
A: For tree-based, it puts factors in order in terms of contribution to variability. You can interpret the report. Look at both and see if they agree.
Q: Is there any course/short videos about mixture DOE + modeling ?
A: https://community.jmp.com/t5/Mastering-JMP/Designing-Mixture-Experiments-Part-1/ta-p/546086 and https://community.jmp.com/t5/Mastering-JMP/Designing-Mixture-Experiments-Part-2/ta-p/549335 At community.jmp.com/mastering you can search for videos on any topic
Q: What tool in the fit model platform should I use when the data distribution is actually a mix of several distributions?
A: In JMP Pro you can use different distributions on your responses. For factors, you may want transform those variables because there is an assumption of normality.
Q: How would you group Housing Price by Low, Mid and High?
A: Use the Utilities>Make Binning Formula. Fit modeling, could you demo?
Resources
- Model Comparison and Selection video and instructions
