\section{Worked example: observation protocol and exact information for the Ehrenfest chain}\label{sec:protocol} Fix $T>0$, $\theta\in[\theta_-,\theta_+]\subset(0,\infty)$, and an integer $\Omega\ge1$. Let $N$ be the stationary chain on $\{0,\ldots,\Omega\}$ with \begin{equation}\label{eq:rates} q_\theta(n,n+1)=\theta(\Omega-n),\qquad q_\theta(n,n-1)=\theta n. \end{equation} The stationary law $\pi_\Omega(n)=2^{-\Omega}\binom\Omega n$ is independent of $\theta$. We observe exact instantaneous counts at $0,\Delta,\ldots,M\Delta$, where \begin{equation}\label{eq:scales} M=\lfloor T/\Delta\rfloor,\quad \tau=M\Delta,\quad t=\theta\Delta,\quad \lambda=\Omega\theta\Delta. \end{equation} Every parameter derivative holds $(\Omega,\Delta,T)$ fixed. In particular, $\lambda$ labels an asymptotic regime and is not a parameter-dependent design. \subsection{Finite-sum likelihood} Each of the $\Omega$ independent particles flips between two states at rate $\theta$. Set $p=(1-e^{-2t})/2$. Conditional on $N_0=n$, the increment is $U-V$, where $U\sim\Bin(\Omega-n,p)$ and $V\sim\Bin(n,p)$ independently. Writing $d=m-n$ gives \begin{equation}\label{eq:transition} P_t(n,m)=\sum_v\binom{\Omega-n}{v+d}\binom nv p^{2v+d}(1-p)^{\Omega-2v-d}, \end{equation} with the sum restricted to valid binomial indices. \begin{proposition}[Exact sampled information]\label{prop:exact} The information about $\theta$ in the sampled counts is \begin{equation}\label{eq:exactFI} \FI_{\Omega,\Delta,T}(\theta) =M\Delta^2\sum_{n,m=0}^{\Omega}\pi_\Omega(n) \frac{(\partial_t P_t(n,m))^2}{P_t(n,m)}. \end{equation} If $K$ is the actual number of jumps in one interval, define \begin{equation}\label{eq:R} R_\Omega(\lambda)=\frac{\Var(\E[K\mid N_0,N_\Delta])}{\lambda}. \end{equation} Then \begin{equation}\label{eq:missing} R_\Omega(\lambda)=1-\frac{\E\Var(K\mid N_0,N_\Delta)}{\lambda}, \qquad \frac{\FI_{\Omega,\Delta,T}(\theta)}{\Omega T/\theta} =\frac{\tau}{T}R_\Omega(\lambda). \end{equation} \end{proposition} \begin{proof} The log likelihood is a sum of transition log likelihoods; its initial term has zero score. Each transition score has conditional mean zero given the past. Cross terms therefore vanish, giving \eqref{eq:exactFI} by stationarity. The total escape rate is $\Omega\theta$, independently of the state, so $K\sim\Pois(\lambda)$ independently of $N_0$. Conditional on $K$, the embedded chain and jump times have no further dependence on $\theta$. The complete score is $(K-\lambda)/\theta$. Differentiation under the finite-state Poisson mixture gives the observed score $\E[K-\lambda\mid N_0,N_\Delta]/\theta$. Conditional variance decomposition proves \eqref{eq:missing}. Over $[0,T]$ the full score has variance $\Omega T/\theta$. \end{proof} The transition-score additivity used here is the stationary first-order case of the Markov formula in \cite[Eqs.~(30) and (34)]{Radaelli2023}, with zero initial-law information. It follows from conditional zero scores, not from independence of successive observations. The observed score is the conditional expectation of the complete score, an instance of the missing-information principle \cite{Louis1982,Pollard2013}; here direct differentiation establishes the identity for this finite-state Poisson mixture.