this is for holding javascript data
Antonino Ingargiola edited Burst_Weights_Theory.tex
about 8 years ago
Commit id: 67153e9d68e39636b49a54bb0dd33c5d614fe958
deletions | additions
diff --git a/Burst_Weights_Theory.tex b/Burst_Weights_Theory.tex
index bc1bf8a..0939df8 100644
--- a/Burst_Weights_Theory.tex
+++ b/Burst_Weights_Theory.tex
...
using the burst size as weights and fitting the peak is equivalent to using
the MVUB estimator for $E_p$.
\subsubsection{Simulations}
It is easy to illustrate the results of previous section through simulations.
In particular we generate a static FRET population of bursts extracting sizes
from an exponential (or gamma) distribution and acceptor counts from a binomial
distribution. By repeatedly fitting the population parameter $E_p$ using a
size-weighted and unweighted average, we verified that the former has systematically
lower variance of the latter as predicted by the theory. Note that this result
holds for any arbitrary distribution of burst sizes.
\subsubsection{Experiments}
Figure X show a comparison of a FRET histogram obtained from the same burst
with and without weights. The burst selection is obtained applying a burst size
of threshold of 20 counts, in order to filter the extreme low-end of the burst size
distribution which include background bursts.
The use of size-weighted FRET histograms allows to obtain a minimal with of
the various FRET peaks without the need of using a manually adjusted high
threshold for burst selection. While an increase in the selection threshold
reduces the shot-noise of the residual population it also discards the useful
information of low and medium size bursts. Necessarly, an unweighted histograms
with high selection threshold will exhibit an higher statistical noise
(due to the reduced number of bursts) compared to the size-weighted histogram
obtained using a low threshold.
statistical variance