1

I am working with a questionnaire which is answered by companies with 5 employees or more.

In my research I want to control for the size of a company, which measurement level would it be? The range of this variable would be 5 - ∞. Would this be an interval variable?

kjetil b halvorsen
  • 77,844
vincitus
  • 43

1 Answers1

5

Stevens' typology of variable types isn't written in stone and it has some problems. See my blog post: Nominal, ordinal, interval, ratio: Stevens' typology and some problems with it.

However, this is usually not relevant when the variable is a control variable; you can simply add it to the model. If the count took only a very few levels (e.g. it was always either 5, 6, or 10) then you might consider adding it as a categorical variable. That seems unlikely here. If you are concerned about nonlinear effects, you could try a spline of the count, or some form of optimal scoring (in SAS this is available in PROC TRANSREG) but that may not be necessary.

Peter Flom
  • 119,535
  • 36
  • 175
  • 383
  • 2
    Very nice references in that post. I can't believe I don't have your blog feed in my reader! – shadowtalker Mar 01 '15 at 14:07
  • 1
    Thank you. I just wanted to know it for the explanation of the variable. I am indeed thinking of transforming it into a categorical variable, but I will decide on this when I will see the distribution of the variable.

    Once again thank you for your clear answer.

     – vincitus Mar 01 '15 at 14:09