this is for holding javascript data
Julian Schrenk edited time_series_convergence.tex
over 9 years ago
Commit id: eb823f6db48e2dfd32aef64a7877de5e36b53637
deletions | additions
diff --git a/time_series_convergence.tex b/time_series_convergence.tex
index 7b7d9a0..20f0927 100644
--- a/time_series_convergence.tex
+++ b/time_series_convergence.tex
...
\begin{equation}
\langle A_i A_j\rangle = \langle A_i \rangle \langle A_j \rangle = \langle A \rangle^2,
\end{equation}
and the second term in Eq.~\ref{eq:delta1} vanishes. Since we are exploring the basin using a MCMC our samples are far from being independent and a more accurate estimate for the statistical error can be derived under the assumption of fast decay as $|i-j|\rightarrow \infty$. With some algebra
\cite{Ambegaokar_2010,Frenkel_2002} \cite{Ambegaokar_2010, Frenkel_2002} it can be shown that the second term on the RHS of Eq.~\ref{eq:delta1} is equal to
\begin{equation}
\label{eq:delta2}
\frac{1}{M^2}\sum_{i \neq j = 1}^M (\langle A_i A_j\rangle = \langle A \rangle^2) \approx \frac{2}{M}\sum_{t = 1}^{\infty}(\langle A_1 A_{1+t}\rangle- \langle A\rangle^2) \equiv \frac{2}{M} \text{Var}A \tau_A,
...
\label{eq:delta3}
\Delta_A^2 = \frac{\text{Var}A}{M} (1 + 2 \tau_A)
\end{equation}
where $\xi_A = 1 + 2 \tau_A$ is the \emph{statistical inefficiency} and $M/\xi_A < M$ is the effective number of uncorrelated samples. We can compute the integrated autocorrelation time and hence the statistical inefficiency using the \texttt{pymbar}
library. library \cite{Chodera_2007, Shirts_2008}.
We can rearrange Eq.~(\ref{eq:delta3}) to find
\begin{equation}