deletions | additions       diff --git a/In_fact_implicit_in_the__.tex b/In_fact_implicit_in_the__.tex  index d4219ad..18cc969 100644  --- a/In_fact_implicit_in_the__.tex  +++ b/In_fact_implicit_in_the__.tex 
...
In fact, implicit in the formulation described here is the idea that the detection scheme used to recover devils is 100\% efficient and any devil encounter with the right geometry will be recovered from the data stream. However, any real detection scheme will have biases and thereby miss some dust devils or skew the recovered profile parameters. In the case of the detection scheme from \citet{Ellehoj_2010}, the choice of 20-s windows means dust devil signals significantly wider than 20-s will not register. register (in fact, none wider than 70-s are reported).  Since it involves smoothing data over 1,000 s windows, the detection scheme from \citet{Jackson_2015} will likewise fail to recover devils as wide or  wider devils. than that.  The inefficiency of the detection scheme can be incorporated directly into our formulation here, though. Instead of recovering the actual underlying distribution of devil parameters, applying Equations \ref{eqn:n-Pact_from_n-Pobs} and \ref{eqn:n-Gammaact_from_n-Gammaobs} will recover those distributions, moderated by the biases of the detection scheme -- $n_{\rm biased} (P_{\rm act})$ and $n_{\rm biased} (\Gamma_{\rm act})$. Taking the next step and converting these distributions into distributions of the actual parameters requires a detailed knowledge of the detection scheme biases. We biases, and we  discuss how to develop this knowledge in Section \ref{sec:discussion_and_conclusions}.