this is for holding javascript data
Antonino Ingargiola edited Burst_Weights_Theory.tex
about 8 years ago
Commit id: f42a08bfc63a0fe34057528b4cc99b61d0c2b395
deletions | additions
diff --git a/Burst_Weights_Theory.tex b/Burst_Weights_Theory.tex
index e78f5f6..68bb230 100644
--- a/Burst_Weights_Theory.tex
+++ b/Burst_Weights_Theory.tex
...
\operatorname{Var} (E_i) = \frac{E_p\,(1 - E_p)}{n_{ti}}
\end{equation}
Bursts with higher
counts carries a counts, provide more accurate estimations
of the population
PR PR, since their
PR variance
will be smaller. is smaller (eq.~\ref{eq:E_var}).
Therefore, in estimating the population PR we need to "focus"
on bigger bursts.
Traditionally, this is accomplished by merely discarding bursts
...
When the statistics $\hat{p}$ is an unbiased estimator of a distribution
parameter and the equality holds in eq.~\ref{eq:cramer_rao},
the estimator is a minimum-variance unbiased
estimator (MVUB)
and the estimator
and it is said to be efficient (meaning that it does an
optimal use the information contained in the sample to estimate the
parameter).
A population of $N$ bursts can be modeled by a set of $N$ binomial
variables with same success probability $E_p$ and varying number of successes
equal to the burst size.
Now, notice An estimator for $E_p$ can be constructed
noticing that the sum of binomial RV with same
success probability is still a binomial (with number of trial equal to
the sum of the number of trials).
Therefore we can build an estimator for $E_p$ by computing the proportion of
success from Taking the sum
over all bursts of acceptor counts
o over all bursts divided by the total
number of photons as in
eq.~\ref{eq:E_estim}. eq.~\ref{eq:E_estim}, we obtain
the proportion of successes $\hat{E}$
\begin{equation}
\label{eq:E_estim}
\hat{E} = \frac{\sum_i n_{ai}}{\sum_i n_{ti}}
\end{equation}
The variance of
$\hat{E}$ (eq.~\ref{eq:E_variance}) is equal to the inverse of
the Fisher information $\mathcal{I}(\hat{E})$ and therefore $\hat{E}$ is a MVUB
estimator for $E_p$.