Specify the prior probabilities and misclassification costs for CART® Classification

Stat > Predictive Analytics > CART® Classification > Prior/Cost

Prior probabilities for response levels

A prior probability is the probability that an observation will fall into a group before you collect the data. When you do not specify prior probabilities, Minitab assumes that the groups are equally likely.
  • All levels have equal probability: The default uses equal prior probabilities for all response levels. For example, if the response has 4 levels, each level is set to 0.25.
  • Probabilities match total sample frequencies: Set prior probabilities according to the sample proportions. For example, if the total number of observations is 1000, and 250 are level 1, 475 are level 2, 100 are level 3, and 175 are level 4, then Minitab uses the proportions of 0.25, 0.475, 0.10, and 0.175 for the levels.
  • Specify a prior probability for each level: Set prior probabilities for each level. Each value must be between 0 and 1. The probabilities for all levels must sum to 1.

Misclassification costs

Specify costs for misclassification. By default, Minitab uses equal costs of 1. To indicate higher costs, use larger values. Costs must be greater than 0.

For example, the following table reflects that the cost of misclassifying a potential customer is 10 times more costly than misclassifying an uninterested customer. The ratio of the costs is relevant, not the actual costs.

Predicted Level
Actual Level Yes (Event) No
Yes (Event)   10
No 1  

By using this site you agree to the use of cookies for analytics and personalized content.  Read our policy