Hello, I am interested in performing a survival analysis with gene expression data. The input would be expression data from a panel of genes, and I want to identify the group (subset) of genes with best model performance. Can JMP (or JMP Genomics) perform an automatic identification of the best model?
I think a question I would first ask is, how would you identify the "best" model? In Both JMP Pro and JMP Genomics there are Validation Methods to help identify the "Best" Model (AICc, BIC, Validation Column, Cross Validation, etc). And then there are fit statistics that also help identify the best performing model method.
As mentioned earlier, JMP Pro does a pretty good job with automatic model selection (number of predictors to use for the model) given the validation method chosen for the model of interest. One still has to compare modeling methods as well as the models themselves. Both JMP Pro and JMP Genomics provide Model Comparisons tools to assess this.
In the end, one should still explore what the genes are and what the model looks like before using. Modeling is an exploratory process and should not be completely "automated".
JMP Genomics info about Predictive Modeling can be found here:
Look into the Buckley-James Estimation method that allows one to use other modeling methods (like Ridge Regression, etc) for Time to Event data with a Censor.
There are no labels assigned to this post.