The sample data table Mushroom.jmp contains 22 characteristics of 8,124 edible and poisonous mushroom. To see probabilities, select Display Options > Show Split Prob from the top red triangle menu.
(a) Use the Partition platform to build a seven split model to predict whether a mushroom is edible based on the 22 characteristics.
(b) Which characteristics are most important? Note: Use Column Contributions.
(c) What are the characteristics of edible mushrooms? Note: Use the Small Tree View and Leaf Report.
(d) Prune back to two splits. What is the predicted probability that a mushroom with no odor and a white spore print color is edible?
(e) Return to the Partition launch window, enter the value 0.3 in the Validation Portion field, and then rerun the model. We use a holdout validation set to determine the best number of splits. Do the validation statistics improve much after three splits? After four splits?
