BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Errors on Errors: Refining Statistical Analyses for Particle Physi
cs
DTSTART;VALUE=DATE-TIME:20190212T151500Z
DTEND;VALUE=DATE-TIME:20190212T171500Z
DTSTAMP;VALUE=DATE-TIME:20190219T185207Z
UID:indico-event-4261@indico.ph.tum.de
DESCRIPTION:\n \n \n \n In a statistical analysis in Particle Physics
\, one faces two distinct challenges: the limited number of particle colli
sions and imperfections in the modelling of the events produced. Roughly s
peaking these correspond to "statistical" and "systematic" errors in the r
esult. To help combat the systematic uncertainties one can try to improve
the statistical model by including additional (nuisance) parameters. The b
est estimate of such a parameter is often treated as a Gaussian distribute
d variable with a given standard deviation. The appropriate values for the
se standard deviations are\, however\, often the subject of heated argumen
t\, which is to say that the errors themselves have errors. A type of mode
l is presented where the uncertainty in the assigned systematic errors is
taken into account. Estimates of the systematic variances are modeled as g
amma distributed variables. The resulting confidence intervals show intere
sting and useful properties. For example\, when averaging measurements to
estimate their mean\, the size of the confidence interval increases as a f
or decreasing goodness-of-fit\, and averages have reduced sensitivity to o
utliers. The basic properties of the model are presented and several types
of examples relevant for Particle Physics are explored.\n \n \n \n\n\n
https://indico.ph.tum.de/event/4261/
LOCATION:Max Planck Institute for Physics Main Auditorium
URL:https://indico.ph.tum.de/event/4261/
END:VEVENT
END:VCALENDAR