sabaruch

Then a first step would be to ahead and look at the skew and kurtosis figures. If it's normally distributed, it should have 0 skew and a kurtosis of about 3. Depending how different these figures are from that, it may or may not be necessary to go through the trouble of figuring out another distribution. For zero skew and high kurtosis, some people use a t-distribution with a smaller number of degrees of freedom.

Part of the problem is that it's often difficult to figure out what distribution you ought to be using. Usually it's the underlying model of probability that determines what distribution you expect... just throwing in a bunch of distributions and seeing which one fits best is effectively data-mining and often just as problematic as assuming something is normally distributed. Which distribution to use generally comes out of your theoretical understanding of what's going on - the parameters are what you determine from empirical sources.

Remember that if you are doing stock returns, you'll want to be looking at ln(total return) when you are doing the fit.