Is "year" as discrete or continuous variable..?

is it proper to use linear regression with "year" (every year from 2009 to 2014) on the x axis and "percent" on the y axis..?

Simple question from a beginner..

  • $\begingroup$ Such a model would almost certainly yield nonsense predictions (percentages outside of [0, 100]) when given years far in the future. Does that matter for your application? $\endgroup$
    – Adrian
    May 24, 2015 at 16:58
  • $\begingroup$ No I don't think so. I just want to show downward trend in students' attendance in a university class from 2009 to 2014. Every class has about 70 - 100 students' in all. $\endgroup$
    – schvost
    May 24, 2015 at 17:06

1 Answer 1


Perhaps you would like to try a transformation of the dependent variable. A useful transformation for your case, might be the logit. That is, your new response would be $Y^{\prime}=\log\frac{Y}{1-Y}$. The new variable will not be bounded and you can apply the usual regression techniques. Assuming of course, the original percentages are not close to either zero or one. If they are, then the transformation will return infinite values.

  • $\begingroup$ Thank you for commenting! No, the percentual values vary from 40% to 80% at the very most. Can I interpret your replies meaning that y = percent and x = years (2009, 2010...2014) would be erroneous? My aim is just to show a trend and whether the trend is significant or not. $\endgroup$
    – schvost
    May 24, 2015 at 18:32
  • $\begingroup$ Or would Poisson regressoin would be an alternative..? $\endgroup$
    – schvost
    May 24, 2015 at 18:34
  • $\begingroup$ @schvost A Poisson regression is used to model count data and therefore is inappropriate here. Your approach is a little crude but it might work too. $\endgroup$
    – JohnK
    May 24, 2015 at 18:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.