I want to sum multiple distributions together accounting for their correlation. What is the most efficient way to do that in python? Unfortunately numpy convolve function seems to be only for independent variables.
Many thanks,
Pierre
Asked
Active
Viewed 116 times
0
-
1If RVs are not independent, then you cannot write joint PDF as product of two PDFs and use convolution theorem to compute it. What is left is just integration, you could use Python `quad` function or similar to compute result. – Severin Pappadeux Jun 08 '20 at 14:51
-
Well, at least in the case of jointly Gaussian variables, I'm pretty sure that something is known about linear combinations or maybe matrix-vector products where the matrix is some known factors and the vector comprises the Gaussian variables. Stuff like that is covered in Chapter VIII of von Mises, Mathematical Theory of Probability and Statistics, and many other places, no doubt. Although this is very interesting, this question is off topic for SO, OP should try stats.stackexchange.com instead. – Robert Dodier Jun 08 '20 at 20:35