problem in using copulafit

I use 'copularnd' function to generate 325 correlated samples in t copulas with linear correlation parameters 0.4 and degrees of freedom 2. When I used 'copulafit' function to fit t copula to those generated samples, the best fitted t copula which is resulted was too different. can any one help me on this? How can I improve the fitting results without any knowledge about dependence structure of the samples?

Answer 1

[Brendan Hamm]

This is a maximum likelihood estimate given a sample so there will always be some uncertainty in the answer. In your case you are using a sample size of only 325. If you need your result to be "more accurate" consider a much larger sample size. Although, I suspect you are not trying to fit parameters for simulated data with a known distribution.