layout | mathjax | author | affiliation | e_mail | date | title | chapter | section | topic | theorem | sources | proof_id | shortcut | username |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
proof |
true |
Joram Soch |
BCCN Berlin |
joram.soch@bccn-berlin.de |
2023-12-08 07:14:47 -0800 |
Maximum-a-posteriori estimation for multinomial observations |
Statistical Models |
Count data |
Multinomial observations |
Maximum-a-posteriori estimation |
P428 |
mult-map |
JoramSoch |
Theorem: Let
Moreover, assume a Dirichlet prior distribution over the model parameter
Then, the maximum-a-posteriori estimates of
$$ \label{eq:Mult-MAP} \hat{p}\mathrm{MAP} = \frac{\alpha_0+y-1}{\sum{j=1}^k \alpha_{0j} + n - k} ; . $$
Proof: Given the prior distribution in \eqref{eq:Mult-prior}, the posterior distribution for multinomial observations is also a Dirichlet distribution
where the posterior hyperparameters are equal to
The mode of the Dirichlet distribution is given by:
Applying \eqref{eq:Dir-mode} to \eqref{eq:Mult-post} with \eqref{eq:Mult-post-par}, the maximum-a-posteriori estimates of
$$ \label{eq:Mult-MAP-s1} \begin{split} \hat{p}{i,\mathrm{MAP}} &= \frac{\alpha{ni} - 1}{\sum_j \alpha_{nj} - k} \ &\overset{\eqref{eq:Mult-post-par}}{=} \frac{\alpha_{0i} + y_i - 1}{\sum_j (\alpha_{0j} + y_j) - k} \ &= \frac{\alpha_{0i} + y_i - 1}{\sum_j \alpha_{0j} + \sum_j y_j - k} ; . \end{split} $$
Since
$$ \label{eq:Mult-MAP-s2} \hat{p}{i,\mathrm{MAP}} = \frac{\alpha{0i} + y_i - 1}{\sum_j \alpha_{0j} + n - k} $$
which, using the
$$ \label{eq:Mult-MAP-qed} \hat{p}\mathrm{MAP} = \frac{\alpha_0+y-1}{\sum{j=1}^k \alpha_{0j} + n - k} ; . $$