I'm trying to test some data for the best distribution fit, and am looking to try the t-dist (all theoretically, no R computations) I think the best way to do this is to assume the distribution is true, use MLE or MoM to estimate parameters and then conduct a goodness-of-fit test. Firstly, is this an appropriate way to do it? Secondly, what is the theoretical derivation for the parameters of the t-dist (using MLE or Mom) And thirdly, how do you estimate the degrees of freedom for a t-test?


  • 1
    $\begingroup$ If you are fitting a t distribution the number of degrees of freedom is a parameter being estimated; it is not given independently. $\endgroup$
    – Nick Cox
    Feb 12, 2014 at 16:05
  • $\begingroup$ Take a look: stats.stackexchange.com/questions/63647/… $\endgroup$
    – Zen
    Feb 12, 2014 at 16:33