3
$\begingroup$

enter image description here

I am struggling with figuring out whether the variable "Education in years" should be used as an ordinal variable or as a continuous variable, or would either work? I haven't coded the data myself.

What I would like to do with the 'education' variable is to run a linear regression with it as a dependent variable. My independent variables are gender, age, occupation etc. Would this be possible with the education data I have? What would you suggest?

$\endgroup$
3
$\begingroup$

The first thing I would do is check out how the variable was coded, even though you didn't do it yourself, and to check out if education in actual years is available. I also wonder what 1 year of education means (probably it means "less than high school" but ... check it out to be sure).

Next, I'm wondering about some of your proposed independent variables. Age and gender make sense, but education can't be dependent on occupation (unless you mean parents' education).

As to your actual question, I think education here is best treated as ordinal or maybe even multinomial. There are various models for ordinal dependent variables, but by far the most common is ordinal logistic regression which depends on the assumption of proportional odds. That seems likely to be violated.

$\endgroup$
  • 3
    $\begingroup$ (+1) But it does make sense to use a regression model purely to predict how long someone's spent in formal education from their job. Also, a continuation ratio ordinal logistic model rather than a proportional odds one might be attractive given that people pass through preceding educational stages to reach the next one. $\endgroup$ – Scortchi - Reinstate Monica Aug 26 '17 at 12:24

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.