11-24-2016 11:04 AM
Assumed that I have datasets, the historical return of stocks for example, and I want to fit the distribution of those returns.
How to approach that using MCMC procedure?
Or any others way can do the distribution fitting?(I tried Severity procedure but worked not good)
Thank you very much!
11-24-2016 01:31 PM - edited 11-24-2016 01:33 PM
proc univariate and proc kde. Or proc fmm for mixtures.
11-25-2016 06:57 PM
You've asked the One Million Dollar Question. There have been countles papers and books written about how to simulate data. I recommend Simulating Data with SAS for a SAS-oriented programming approach. If it is univariate data, PROC UNIVARIATE fits many common distributions. However, in many cases you need to ask yourself "what features of this data are important" and then make sure the simulation preserves those features.