Economics Job Market Rumors » Econometrics Job Rumors

ordinal independent variables

(8 posts)
  • Started 4 months ago by Anonymous
  • Latest reply from anonymous
  1. Anonymous
    Unregistered

    If you have independent variables that are ordinal (say 0, 1, 2, 3) with higher values denoting stronger degree of the variable then do you just include them in a regression as dummies? or treat them as continuous?
    would this differ if the independent variable was education {none, primary, secondary}
    thanks

    Posted 4 months ago #
  2. Anonymous
    Unregistered

    Include them as dummies; do not treat them as continuous, since doing so means that every "step" measures the same thing (e.g., the move from no education to primary education is identical to the move from primary education to secondary education), which is rarely true in the case of ordinal variables.

    Posted 4 months ago #
  3. Anonymous
    Unregistered

    thanks. what if the variable measures something as low, medium and high or none, some, all?

    Posted 4 months ago #
  4. Anonymous
    Unregistered

    ^Same as in the education case. Even if it measured equal intervals, you'd be throwing away information by imposing a homogeneous impact. Better allow for heterogeneous impacts between discrete categories, which is more flexible anyway.

    Posted 4 months ago #
  5. Anonymous
    Unregistered

    thanks again. very helpful

    Posted 4 months ago #
  6. Anonymous
    Unregistered

    lol

    Posted 3 months ago #
  7. Anonymous
    Unregistered

    no kidding

    Posted 3 months ago #
  8. Anonymous
    Unregistered

    reverse ordered probit

    Posted 3 months ago #

RSS feed for this topic

Reply

(required)

Allowed markup: a blockquote code em strong ul ol li.
You can also put code in between backtick ( ` ) characters.