0

A CDF of a performance task

Hi all, what kind of functional form of CDF do you think closely resembles this shape? Thanks!

  • Pick a software, say Mathamtica, or R and choose a family of likely families. Fit your data within each family using the method of maximum likelihood, and then compare values of maximal log-likelihoods and pick the family that yields the highest likelihood value. That would be your best bet. – Sasha Sep 05 '13 at 13:45
  • What is the source of your data? – gammatester Sep 05 '13 at 13:47
  • The source of my data is from an experiment I conducted. I was trying to build a simple theoretical model that could aproximate the outcomes from my experiment. I asked a question here link regarding that – user91909 Sep 05 '13 at 13:56
  • Thanks Sasha, I unfortunately have never used such software, can you provide a link to a good guide for R? Thanks – user91909 Sep 05 '13 at 13:57

1 Answers1

0

Since the "score" seems to be a non-negative integer variable, and the CDF is S-shaped, a binomial distribution (http://en.wikipedia.org/wiki/Binomial_distribution) might have a good fit.

Another candidate (with just one free parameter) is the Poisson distribution (http://en.wikipedia.org/wiki/Poisson_distribution).

In both cases, it should be straightforward to fit the distribution to your data using maximum likelihood estimation.

Martin
  • 475