talked at length about how to implement the EW uncertainties and upper limits with pat, and from pat’s SM batch file seems pretty easy. here’s the notes:
EW uncertainties and upper limits…how to deal:
First have to take the 900 simulated EWb values and plot them as a CDF. The observed EWb that we are trying to match to will have some PDF described by a Gaussian where mu=obsEWb and sigma=obsError. The cumulative distribution function of this PDF will be an error function. If we map the y-axis of these two graphs onto a third graph, such that the Gaussian value is assosciated with the x-axis and the CDF value is associated with the y-axis, we generate a function similar to the error function which describes how well the simulated values matches the observed.
In the analysis in the manuscript, we approximate the input erf (x) as step function, and this only one point is mapped to the output, basically one number. Now that we are using the input erf(x), we are generating a distribution of probabilities. Each of the 112 sets of data will have one of these output distributions, that must then be used to generate a CDF to compare to y=x.