I saw in some papers that people sometimes use empirical cdf or complementary cdf to quantify the distribution of data.

What is the advantage of using them and what interesting can they tell about data structure?

  • 2
    $\begingroup$ Advantage: You don't need to find a theoretical distribution that fits your data (this can be really tricky!). Also it might give you an idea of what theoretical distribution fits your data if you need a theoretical distribtions for forecasting or testing or whatever. $\endgroup$ – elevendollar Feb 19 '15 at 16:39
  • 1
    $\begingroup$ It is the only elegant way to describe the distribution of a continuous variable without information loss. $\endgroup$ – Michael M Feb 19 '15 at 17:13
  • 3
    $\begingroup$ "advantage" -- compared to what when doing what? $\endgroup$ – Glen_b -Reinstate Monica Feb 20 '15 at 3:44

As elevendollar mentions the empirical cdf does not require any assumption about functional form. The only parameter(s) you may have to adjust is the bin width if the CDF is determined via histogram, or smoothing width if you are smoothing data, etc..

It is all too often assumed that data fits a commonly used distribution, such as a Gaussian. Parameters are then estimated under the Gaussian assumption, compared, and conclusions drawn, even though there's no a priori reason to believe so.


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.