I am seeking an approach to sampling conditional joint distribution (new to probability). I will put my case in a simple way:

Similar question for discrete variables is asked here but not yet answered; I am interested in continuous random variables.

I have observations/ samples (historical synchronized data of forecasted and actual values corresponding to those forecasts) of 6 random variables, i,e. I have 12 random variables [X1 X2... X6, Y1 Y2... Y6] where Xs are the actual values and Ys are the corresponding forecasted values (matrix of 10807*12). So, I am assuming I have enough samples from the joint distribution of the random vector of 12 random variables.

My target is to find the samples from Conditional Error Distribution (error E1=X1-Y1) of the form as in image of the interested conditional distribution for specific forecast values y1 y2... y6 written in word, sorry for image!

I can fit the joint distribution of random vectors [E1 E2... E6, Y1 Y2... y6] (as values calculated for E are also synchronized to the observed data) and [X1 X2... X6, Y1 Y2... Y6] through linear or rank correlation or copula(s). So, basically, I can generate samples from joint distributions of any combination of the random variables, and also from any marginal distribution too.

My question: How to sample from the above conditional joint distribution when specific values of the forecast are provided; I can only generate samples from any joint combination of actual values, forecasted values, and error values? can this be done?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.