0
$\begingroup$

I need to do a formal report with the results of a decision tree classifier developed in SPSS, but I don't know how. I know there are really well defined ways to report statistics such as mean and standard deviation (e.g., M = 19.22, SD = 3.45) or correlations (e.g., r(55) = .49, p < .01), but I cant find a standard for decision trees. I have statistics such as predicted/observed accuracy percentage, risk estimate (resubstitution and cross validation), and standard error. Do I need other or these are enough? How should I show them?

$\endgroup$

1 Answer 1

1
$\begingroup$

You should report somewhere the full specification you used to build the tree and the (mis)classification table. You might want to report the tree itself or the rule set it generated depending on the audience.

$\endgroup$
3
  • $\begingroup$ I'm building about 15 trees to classify different variables, and some of them are massive (e.g., 300 nodes). I wanted to show them but it became impossible. I'll report the classification results, but should I include risk estimates or cross validation results? $\endgroup$
    – vabm
    Jan 9, 2016 at 11:20
  • $\begingroup$ It really depends on what you want to convey to the reader. A table of all the risks would be compact and perhaps informative, but if you are mainly using these for prediction purposes and have new data, the effectiveness on that might be more interesting to the audience. $\endgroup$
    – JKP
    Jan 10, 2016 at 18:16
  • $\begingroup$ Perfect, just what I needed. I thought there could be an standard for these reports, but as you say I will report what needed depending the case. $\endgroup$
    – vabm
    Jan 10, 2016 at 21:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.