1

When building the regression models, some independent variables are of the following type x={1, 2, -100.1, 200.1, 300, 0, 0, 0, 0, 4000}

If there exists an internal order relationship, should we treat them as ordered variable or continuous variable?

If there does not exist an internal order relationship, should we just treat them as the categorical variable?

Technically, how does R handle these scenarios? Statistically, what are the effects on the model performance by using different treatment?

user785099
  • 1,105
  • 3
  • 14
  • 24
  • 4
    The typology of data refers to kinds of *measurements.* The numbers alone cannot reveal the type (and therefore software has no automatic way to "handle" them--it has to be instructed). What are your variables, then, and how were they measured? – whuber Oct 14 '14 at 23:41
  • 1
    Note that a count variable certainly isn't continuous, but it's not categorical, nor ordered-catgeorical. On Stevens' scale typology, it's usually ratio - interval and ratio aren't necessarily continuous. I'd say that the way we look at a variable is determined (at least in part) by what we're doing with the data. The same set of numbers used to answer one question may be looking at it as nominal, but a different question might see the same data as ordinal, while a third sees it as interval or even ratio (not that there's anything particular special about Stevens' typology - there are others). – Glen_b Oct 15 '14 at 00:07
  • 1
    ... see [here](http://stats.stackexchange.com/questions/106393/should-types-of-data-nominal-ordinal-interval-ratio-really-be-considered-types/106400#106400) for more detail and some references, and also some of the discussion [here](http://stats.stackexchange.com/questions/113737/factor-or-no-factor/113767#113767), which discusses particular examples of how the same variable might be regarded differently depending on what you're using it for. – Glen_b Oct 15 '14 at 00:15

0 Answers0