Every thing you must know concerning the Dirichlet distribution
The Dirichlet distribution is a generalization of the beta distribution. In Bayesian statistics, it’s generally used because the conjugate previous to the multinomial distribution, therefore it may be used to mannequin the uncertainty of a random vector of chances. It has a variety of purposes together with Bayesian evaluation, textual content mining, statistical genetics, and nonparametric inference. This text offers an intuitive introduction to Dirichlet distribution and reveals how it’s related to the multinomial distribution. As well as, it reveals how it may be modeled and visualized in Python.
Definition
Suppose that the continual random variables X₁, X₂, …Xₖ (ok≥2) kind the random vector X outlined as:
We additionally outline the vector α as:
the place
Now the random vector X is alleged to have Dirichlet distribution with parameter α if it has the next joint PDF:
The operate B(α) is named the multivariate beta operate and is outlined as
the place Г(x) is the gamma operate. If the random vector X has a Dirichlet distribution with parameter α, it’s denoted by X ~ Dir(α). The multivariate beta operate is included within the joint PDF to normalize it. The joint PDF ought to combine to 1 over its area:
Therefore, we’ve:
Primarily based on Equation 1, the values that the random variables X₁, X₂, …Xₖ take ought to meet the next circumstances to have fₓ(x)>0:
These circumstances outline the help of the Dirichlet distribution. The help of X, and of its distribution, is the set of all x (the values that X can take) the place fₓ(x)>0. If X has ok components, the help of X with a Dirichlet distribution is a ok-1 dimensional simplex. A simplex is a bounded linear manifold that’s created due to the constraints of Equation 3. A simplex is the generalization of the notion of a triangle to larger dimensions. Therefore, a ok-1…
