The beta distribution is $$ (\text{constant})\times x^{\alpha-1}(1-x)^{\beta-1} \, dx \quad\text{for } 0\le x\le 1. $$ Supposing $X$ to be so distributed, one has \begin{align} & \mu = \operatorname E X = \frac \alpha {\alpha+\beta}, \\[10pt] & \nu = \operatorname E(1-X) = \frac \beta {\alpha+\beta}, \\[4pt] & \text{(so that $\mu+\nu=1$)} \\[10pt] & \operatorname{var}(X) = \operatorname{var}(1-X) = \frac{\mu\nu} \kappa \quad \text{where } \kappa = \alpha+\beta + 1. \end{align} Thus the family of distributions is parametrized by $(\mu,\kappa)$ or by $(\nu,\kappa).$

I have on occasion called $\kappa$ a “concentration parameter.” Is there some standard name for it?


I've heard that $\kappa := \alpha+\beta$ is sometimes also referred to as the precision of the beta distribution, precisely because if $\mu,\nu$ are fixed, then larger $\kappa$ implies lower variance. The term's meaning should carry over to your definition with the extra $+1$.

| cite | improve this answer | |
  • $\begingroup$ It seems odd to use $\alpha+\beta$ rather than $\alpha+\beta+1,$ since if $\alpha,\beta$ both $\to0$ with the expected value fixed at $1/2$, then the variance approaches $1/4$ but $1/(\alpha+\beta) \to+\infty. \qquad$ $\endgroup$ – Michael Hardy Jun 19 '18 at 20:30

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.