I notice that on many examples one is keen to convert Age to a categorical age range.

I am wondering if that is always necessary.

The famous golf play decision tree example has ranges for temperatures up-front. I would have thought from other examples the golf play with actual temperatures would give near enough the same predictions. I am going to test that in the next free time slot.

I am looking at a diabetes case in which the example states let us "bin" the ages. To me the age is an automatically in its own right ordinal attribute and I would have thought tat for logistic regression we would not need to bin / categorize it. Age has a limited range.

  • $\begingroup$ spline it !!!!!!!! $\endgroup$ Jan 6, 2021 at 18:34
  • $\begingroup$ @kjetilbhalvorsen pls elaborate $\endgroup$ Jan 6, 2021 at 18:35
  • 1
    $\begingroup$ This post for instance has examples of splining age $\endgroup$ Jan 6, 2021 at 18:40