1

Usually age is rounded down (i.e $15.7\simeq 15$). So shouldn't we treat it as an ordinal and not a ratio scale variable?

If this is a duplicate question, please refer me to the duplicated question.

Nick Cox
  • 56,404
  • 8
  • 127
  • 185
Ahmed Ali
  • 133
  • 5
  • 3
    Why should the rounding change whether a variable is ordinal or ratio scaled? – Stephan Kolassa Jul 07 '15 at 20:07
  • @StephanKolassa I know that theoretically we can measure age with more and more precision. But if rounding does not affect measurement scale then shouldn't we also treat say young-middle aged-old as a ratio scale variable? – Ahmed Ali Jul 07 '15 at 20:13
  • For example, a pair of people with ages 17.5 and 18.8 will be categorized as 17 and 18 respectively. Another pair of people with ages 18.5 and 19.9 will be categorized as 18 and 19. The difference is 1 in both cases but the age difference is not the same. – Ahmed Ali Jul 07 '15 at 20:17
  • 3
    "Young-middle aged-old" is not quantified. That's why it's not inherently a ratio variable, whereas age represents a meaningful quantity. Having said that, I wish to emphasize that such questions tend to be more of a diversion than helpful: how you treat age in a model should depend on how it is being used and what you want to learn. It's not hard to think of circumstances where the analysis will be improved when age is treated as ordinal or even nominal. – whuber Jul 07 '15 at 20:17
  • @whuber Then should we treat age to the nearest 5 years multiple as a ratio scale? Thanks for both of you for your replies btw. – Ahmed Ali Jul 07 '15 at 20:26
  • 4
    I would give the same answer: it depends on the considerations I mentioned. Age, to the nearest multiple of five, definitely can be used as a ratio variable to good effect in some analyses, such as any analysis that posits a linear relationship between age and another variable. – whuber Jul 07 '15 at 20:29

0 Answers0