Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/external/xmm_sw_cal/icwg/presentations/SystematicErrors.pdf Дата изменения: Fri Jun 16 13:03:02 2006 Дата индексирования: Sat Dec 22 15:02:48 2007 Кодировка: Поисковые слова: южная атлантическая аномалия

Handling Systematic Errors


Basics

Define as errors in modeling of instrument Discuss how to handle systematic errors, not how to generate them
Generating real systematic errors is very difficult Most are merely estimated

Cal scientists asked to quantify systematic errors but quotes may be too simplistic Admonish users to avoid xspec syserr


Systematic Error Effects
parameters can depend on systematic errors in nontrivial ways avoid two types of problems
A: claims of new physics that actually result from known cal errors B: features may be ignored because they are erroneously assumed to be due to poor calibration


Delineate Cases: T e? riag
Simplest: statistical errors dominate ignore systematic errors Easy cases: e.g. line fluxes, ratios, energies Hardest cases: contact PI or project team
consider collaborations e.g.: separating instrumental & source edges

In between: need something more for users


Options for Users
detector physics is poor

Avoid xspec syserr: errors are correlated Simulate full detector (e.g. GEANT) ? No.
instrument parameters inaccurate

Fake it to get lower limits to errors
fit model to obser ved data use model to generate random data fit fake data and estimate parameter errors errors are statistical contribution to true errors can still make A & B errors


Options for Users (cont.)
Generate many response functions (RFs) (see Drake et al. 2005, Chandra Cal W/S)
new functions are consistent within syst. errors fit data; param distribution from syst. errors problem: physically unreasonable RFs cal scientists should test method problem: consistency of adjustments (As) problem: As can mask or fake new physics user As should be posted or sent to cal group if similar RF As found, cal fixes RFs

Allow users to adjust RFs within syst. errors


Other

Consider sof tware errors
post results from standard reductions admonish users to avoid xspec syserr

include errors in background model
cal provides particle flux predictions cal provides spatial, spectral dependencies

Missions should include "watch-out" or caveats web pages


Actions

Cal scientists should

admonish users to avoid xspec syserr try Drake et al. method publish algorithm for calc'ing statistical errors obtain/post user's adjustments and feedback include "watch-out" or caveats web pages post results from standard reductions provide background models, usage details give examples of simple handling of syst. errors

Missions should