Systematic Astro Statistics

1
Systematic Astro Statistics
Vinay Kashyap CHASC
Hyunsook Lee, Jeremy Drake, Pete Ratzlaff, David
van Dyk, Rima Izem, Xiao-li Meng, Taeyoung
Park, Aneta Siemiginowska, Alanna Connors,
Andreas Zezas
2
CHASC
CHASC AstroStatistics Collaboration
c.1997 Chandra X-ray Center and Harvard
University AstroStatistics Collaboration
Chandra AstroStatistics Cabal

California-Harvard AstroStatistics Collaboration

Members of the University are expected to take
individual responsibility for their participation
in any outside activity, and use their best
efforts to avoid false or misleading suggestions
by others that the activity is an undertaking of
Harvard or any of its units. In general, all
members should observe the Universitys policy on
the use of the Harvard name, and limit their
identification with Harvard to listing their
formal titles as appropriate. CHASC is not an
official entity recognized by Harvard University
(HU). HU does not endorse any program listed
herein. Reference to any product,
process, publication, service, or offering of any
third party by trade name, trademark,
Manufacturer or otherwise does not necessarily
constitute or imply the endorsement or
recommendation of such by HU. HU is not
responsible for representations made here, and
does not explicitly or implicitly endorse the
views presented herein. And so on and so forth,
and so on and
3
CHASC
high-energy, low-counts, Poisson workshops,
papers, codes, weblog
4
CHASC webpage http//hea-www.harvard.edu/AstroSta
t/
The AstroStat Slog http//groundtruth.info/AstroS
tat/slog/
5
AstroStat Slog visibility Apr 10 - Apr 30, 2008
6
Project Incorporate Systematic Error In X-ray
Data Analysis
The three most important effects that affect data
analysis Calibration! Calibration! Calibration!
7
X-ray analysis from model to data
van Dyk et al. 2001, ApJ, 548, 224
8
from there to here
(E,p,t) photon energy, location, arrival
time (E,p) detector channel, chip location S
astrophysical source model R energy
redistribution function (RMF) P position
redistribution function (PSF) A effective area
(ARF) M predicted model counts
9
uncertainty in effective area
Drake, Ratzlaff, et al. (2005)
10
uncertainty in energy response
11
effect on model parameter uncertainty
Probability density
?
12
but how exactly?
13
MCMC
14
DATA
CALIBRATION
Draw parameters
Compute likelihood
Update parameters
15
DATA
CALIBRATION
Draw effective areas
Draw parameters
Compute likelihood
Update parameters
16
but where do the effective areas come from?
17
Principal Components
18
eigenvector
eigenvalue
N (0,1)
new
nominal
bias
PCs
residual
A A0 bias components residual
19
what have we got so far?

• realistic error bars
• implemented in BLoCXS
• Sherpa
• 100x drop in storage
• 500x speed up in analysis
• generalize to any instrument
• extendable to model lacunae
• roll your own

20
whats next?
unified file format for ingestion into
XSPEC/Sherpa other schemes of dimensionality
reduction RMFs 2D PCA, within and between
PCA PSFs multiscale residuals
21
effect on parameter uncertainty