Dear all To generate scenarios for stochastic programming models, random samples are drawn from the probability distribution function of stochastic variables and the variable values and their probabilities are recorded. With this approach, there is no reason that sum of the probability of the scenarios become unity. Is it needed any postprocessing on the probability of the generated scenarios? Regards; Hossein asked 27 Jan '15, 09:53 Hossein 
If you are sampling scenarios from a probability distribution, then the resulting scenarios are usually considered to be equally likely. If you are somehow splitting the probability distribution to generate different scenarios, then this splitting gives you the right probabilities for the scenarios. answered 01 Feb '15, 05:02 JF Meier Dear JF Meier Many thanks for your clear comment. The problem variables are highly correlated. I am interested to sample based on their conditional probabilities. I know that correlation and conditional probability are different subjects. I can apply correlation using Cholesky factorization. But, I want to know 1) Is there any method to generate the samples based on the conditional probabilities. I have many historical data but I do not know how the conditional probabilities (or joint distribution) can be made using these data. 2) In general, is there a mathematical formulation to relate the correlation to the conditional probabilities? Would you please guide me? Thanks in advance; Regards; Hossein
(01 Feb '15, 06:54)
Hossein
