this is for holding javascript data
Brian Jackson edited Ultimately_though_we_re_interested__.tex
over 8 years ago
Commit id: 260deedfdfea81c0acba605593f11c914c5bc51b
deletions | additions
diff --git a/Ultimately_though_we_re_interested__.tex b/Ultimately_though_we_re_interested__.tex
index aad7fd3..1b3fffb 100644
--- a/Ultimately_though_we_re_interested__.tex
+++ b/Ultimately_though_we_re_interested__.tex
...
Ultimately,
though, we're we are interested in
working from converting the
density of observed
distribution back to the
underlying distribution. density of actual parameters. Fortunately,
the combined recovery bias and distortion expression Equation \ref{eqn:convert_from_actual_to_observed_density} provides a simple way
of working from the observed to
do this:
\begin{equation}\label{eqn:n-Pact_from_n-Pobs}
n(P_{\rm act}) = -P_{\rm act}^{-1} \left( \dfrac{P_{\rm max} - the actual distribution of physical parameters. Differentiating with respect to $b(\Gamma_{\rm obs}, P_{\rm
min}}{P_{\rm min}} \right) \left( \frac{d}{dP_{\rm obs}} \right)_{P_{\rm obs} = obs})$ and then solving for $\rho(\Gamma_{\rm act}(b), P_{\rm
act}} \bigg[ n(P_{\rm obs})\ act}(b)) gives
\begin{equation}
\label{eqn:convert_from_observed_to_actual_density}
\rho(\Gamma_{\rm act}(b), P_{\rm act}(b)) = \frac{1}{2} b_{\rm max}^2 f(b) \dfrac{d\rho(\Gamma_{\rm obs}, P_{\rm
obs}^2 \bigg], obs}}{db}
\end{equation}
where $\left( \frac{d}{dP_{\rm obs}} \right)_{P_{\rm obs} = P_{\rm act}}$ means to calculate the $P_{\rm obs}$-derivative of the expression in square brackets and then replace $P_{\rm obs}$ with $P_{\rm act}$.
It's clear that plugging Equation \ref{eqn:n-Pobs_from_uniform_n-Pact} into Equation \ref{eqn:n-Pact_from_n-Pobs} recovers the original uniform distribution. On the surface, though, Equation \ref{eqn:n-Pact_from_n-Pobs} suggests the strange result that for the power-law distribution $n(P_{\rm obs}) \sim P_{\rm obs}^{-2}$, similar to that reported in \citet{Jackson_2015}, the underlying distribution $n(P_{\rm obs}) = 0$. However, the approach here assumes that $P_{\rm obs}$-values only span a finite range, which is violated by the simple $P_{\rm obs}^{-2}$ power-law. Instead, a form similar to Equation \ref{eqn:n-Pobs_from_uniform_n-Pact} can be used to describe a power-law, with the 2 replaced by the desired index.