How can I compute a confidence interval for a predicted value computed using EVAL command and a classification and regression tree in Statistics Toolbox 6.1 (R2007b)?

I generate a classification and regression tree and evaluate a new data point with EVAL command:

load fisheriris;

t = classregtree(meas,species,'names',{'SL' 'SW' 'PL' 'PW'});

sfit=eval(t,meas(1,:)+rand(1,4))

I would like to compute the confidence level for the variable 'sfit'.

Answer 1

[MathWorks Support Team]

There is no command available in Statistics Toolbox in MATLAB to calculate the confidence interval for a predicted response value given a new data point and a classification and regression tree, because Classification and Regression Trees (CART) technique is not based on a probabilistic model. There is no probability level or confidence interval associated with predictions derived from using a CART tree to classify a new set of data. The confidence that an analyst can have in the accuracy of the results produced by a given model (that is, a tree) is based purely on its historical accuracy—how well it has predicted the desired response in other, similar circumstances. For more information, consult, for example, the following introductory reference on CART:

 <http://www.ifpri.org/themes/mp18/techguid/tg03.pdf>