0

I have the following variables that follow a beta distribution:

           Alpha1     Alpha2     Beta(Alpha1,Alpha2)
Variable 1     3.18       2.81       0.53
Variable 2     2.33       3.666      0.38
Variable 3     1.44       4.55       0.24
Variable 4     1.8        4.2        0.3
Variable 5     2.61       3.38       0.436

From the distribution I have generated random values in order to obtain the Pearson correlation matrix. The problem is, in general, random numbers of univariate variables, the correlation is very low. My question is this:

From the beta functions that I show above how you could generate a multivariable Dirichlet function in python and how you could obtain its correlation matrix.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
turaran32
  • 1
  • 3
    Of course the correlation is low--you are generating uncorrelated values. Is this your intention or not? – whuber Jul 24 '21 at 13:52
  • In theory, they are values that present a degree of correlation, so I want to study the variables as multivariables, but I don't really know how to go from the beta function to the Dirichlet function with the data I have. – turaran32 Jul 25 '21 at 16:11
  • 1
    If by "in theory" you mean *in practice,* then evidently you need a different model for your variables, because these aren't correlated. Please consider telling us the problem you actually face, because the answers are likely to be more useful to you. – whuber Jul 25 '21 at 17:05
  • The problem I have is that I do not know how to obtain the alphas of the dirichelt distribution from the univariable beta functions that I show above. – turaran32 Jul 25 '21 at 17:31
  • [Use Gamma distributions](https://stats.stackexchange.com/questions/36093). You can [use Beta distributions to construct a generalized Dirichlet distribution](https://stats.stackexchange.com/questions/534411). – whuber Jul 25 '21 at 18:38

0 Answers0