Index: The Book of Statistical ProofsStatistical Models ▷ Count data ▷ Poisson distribution with exposure values ▷ Log model evidence

Theorem: Consider data $y = \left\lbrace y_1, \ldots, y_n \right\rbrace$ following a Poisson distribution with exposure values:

\[\label{eq:Poiss-exp} y_i \sim \mathrm{Poiss}(\lambda x_i), \quad i = 1, \ldots, n \; .\]

Moreover, assume a gamma prior distribution over the model parameter $\lambda$:

\[\label{eq:Poiss-exp-prior} p(\lambda) = \mathrm{Gam}(\lambda; a_0, b_0) \; .\]

Then, the log model evidence for this model is

\[\label{eq:Poiss-exp-LME} \begin{split} \log p(y|m) = &\sum_{i=1}^n y_i \log(x_i) - \sum_{i=1}^n \log y_i ! + \\ &\log \Gamma(a_n) - \log \Gamma(a_0) + a_0 \log b_0 - a_n \log b_n \; . \end{split}\]

where the posterior hyperparameters are given by

\[\label{eq:Poiss-exp-post-par} \begin{split} a_n &= a_0 + n \bar{y} \\ a_n &= a_0 + n \bar{x} \; . \end{split}\]

Proof: With the probability mass function of the Poisson distribution, the likelihood function for each observation implied by \eqref{eq:Poiss-exp} is given by

\[\label{eq:Poiss-exp-LF-s1} p(y_i|\lambda) = \mathrm{Poiss}(y_i; \lambda x_i) = \frac{(\lambda x_i)^{y_i} \cdot \exp\left[-\lambda x_i\right]}{y_i !}\]

and because observations are independent, the likelihood function for all observations is the product of the individual ones:

\[\label{eq:Poiss-exp-LF-s2} p(y|\lambda) = \prod_{i=1}^n p(y_i|\lambda) = \prod_{i=1}^n \frac{(\lambda x_i)^{y_i} \cdot \exp\left[-\lambda x_i\right]}{y_i !} \; .\]

Combining the likelihood function \eqref{eq:Poiss-exp-LF-s2} with the prior distribution \eqref{eq:Poiss-exp-prior}, the joint likelihood of the model is given by

\[\label{eq:Poiss-exp-JL-s1} \begin{split} p(y,\lambda) &= p(y|\lambda) \, p(\lambda) \\ &= \prod_{i=1}^n \frac{(\lambda x_i)^{y_i} \cdot \exp\left[-\lambda x_i\right]}{y_i !} \cdot \frac{ {b_0}^{a_0}}{\Gamma(a_0)} \lambda^{a_0-1} \exp[-b_0 \lambda] \; . \end{split}\]

Resolving the product in the joint likelihood, we have

\[\label{eq:Poiss-exp-JL-s2} \begin{split} p(y,\lambda) &= \prod_{i=1}^n \frac{ {x_i}^{y_i}}{y_i !} \prod_{i=1}^n \lambda^{y_i} \prod_{i=1}^n \exp\left[-\lambda x_i\right] \cdot \frac{ {b_0}^{a_0}}{\Gamma(a_0)} \lambda^{a_0-1} \exp[-b_0 \lambda] \\ &= \prod_{i=1}^n \left(\frac{ {x_i}^{y_i}}{y_i !}\right) \lambda^{\sum_{i=1}^n y_i} \exp\left[-\lambda \sum_{i=1}^n x_i\right] \cdot \frac{ {b_0}^{a_0}}{\Gamma(a_0)} \lambda^{a_0-1} \exp[-b_0 \lambda] \\ &= \prod_{i=1}^n \left(\frac{ {x_i}^{y_i}}{y_i !}\right) \lambda^{n \bar{y}} \exp\left[-n \bar{x} \lambda\right] \cdot \frac{ {b_0}^{a_0}}{\Gamma(a_0)} \lambda^{a_0-1} \exp[-b_0 \lambda] \\ &= \prod_{i=1}^n \left(\frac{ {x_i}^{y_i}}{y_i !}\right) \frac{ {b_0}^{a_0}}{\Gamma(a_0)} \cdot \lambda^{a_0 + n \bar{y} - 1} \cdot \exp\left[-(b_0 + n \bar{x}) \lambda\right] \\ \end{split}\]

where $\bar{y}$ and $\bar{x}$ are the means of $y$ and $x$ respectively:

\[\label{eq:xy-mean} \begin{split} \bar{y} &= \frac{1}{n} \sum_{i=1}^n y_i \\ \bar{x} &= \frac{1}{n} \sum_{i=1}^n x_i \; . \end{split}\]

Note that the model evidence is the marginal density of the joint likelihood:

\[\label{eq:Poiss-exp-ME} p(y) = \int p(y,\lambda) \, \mathrm{d}\lambda \; .\]

Setting $a_n = a_0 + n \bar{y}$ and $b_n = b_0 + n \bar{x}$, the joint likelihood can also be written as

\[\label{eq:Poiss-exp-JL-s3} p(y,\lambda) = \prod_{i=1}^n \left(\frac{ {x_i}^{y_i}}{y_i !}\right) \frac{ {b_0}^{a_0}}{\Gamma(a_0)} \frac{\Gamma(a_n)}{ {b_n}^{a_n}} \cdot \frac{ {b_n}^{a_n}}{\Gamma(a_n)} \lambda^{a_n-1} \exp\left[-b_n \lambda\right] \; .\]

Using the probability density function of the gamma distribution, $\lambda$ can now be integrated out easily

\[\label{eq:Poiss-exp-ME-qed} \begin{split} \mathrm{p}(y) &= \int \prod_{i=1}^n \left(\frac{ {x_i}^{y_i}}{y_i !}\right) \frac{ {b_0}^{a_0}}{\Gamma(a_0)} \frac{\Gamma(a_n)}{ {b_n}^{a_n}} \cdot \frac{ {b_n}^{a_n}}{\Gamma(a_n)} \lambda^{a_n-1} \exp\left[-b_n \lambda\right] \, \mathrm{d}\lambda \\ &= \prod_{i=1}^n \left(\frac{ {x_i}^{y_i}}{y_i !}\right) \frac{\Gamma(a_n)}{\Gamma(a_0)} \frac{ {b_0}^{a_0}}{ {b_n}^{a_n}} \int \mathrm{Gam}(\lambda; a_n, b_n) \, \mathrm{d}\lambda \\ &= \prod_{i=1}^n \left(\frac{ {x_i}^{y_i}}{y_i !}\right) \frac{\Gamma(a_n)}{\Gamma(a_0)} \frac{ {b_0}^{a_0}}{ {b_n}^{a_n}} \; , \end{split}\]

such that the log model evidence is shown to be

\[\label{eq:Poiss-exp-LME-qed} \begin{split} \log p(y|m) = &\sum_{i=1}^n y_i \log(x_i) - \sum_{i=1}^n \log y_i ! + \\ &\log \Gamma(a_n) - \log \Gamma(a_0) + a_0 \log b_0 - a_n \log b_n \; . \end{split}\]

Metadata: ID: P43 | shortcut: poissexp-lme | author: JoramSoch | date: 2020-02-04, 15:12.