|
Verification and Uncertainty Quantification
The level of accuracy required by dark energy experiments represents a
major challenge: Predictions and data analysis methods must match --
and preferably exceed -- the observational accuracy. Achieving this
goal requires (i) rigorous code verification, and (ii) advanced
methods for estimating cosmological and model parameters using
observations and simulations. We have recently carried out a
comprehensive code comparison project, finding that statistics such as
the matter power spectrum only agree at the ~5-10% level in regimes
where much better results were expected. (An extended code comparison
project is in progress. More soon!) This is an order of magnitude shy
of the required accuracy. A major new effort is required to discover
the ultimate error limits.
A key challenge for precision cosmology is the accurate determination
of parameters from observations and simulations. Commonly-used fitting
functions are too crude for precision work. Direct use of simulations
is essential for the physics to be correctly captured. Simulations,
however, are very expensive, and only a restricted number can be
performed; a billion-particle simulation with HOT takes 2*10^17 flops
of computation at a cost of 30K processor-hours. A new methodology to
combat this problem has been outlined by us recently and will be a
major cornerstone of this project.
Publications
- Simulations and cosmological inference: A statistical model for power spectra means and covariances
M.D. Schneider, L. Knox, S. Habib, K. Heitmann, D. Higdon, C. Nakhleh,
Phys. Rev. D submitted,
arXiv:0806.1487
- Constraining Cosmological Parameters by Combining Simulations and
Physical Observations
D. Higdon, C. Nakhleh, S. Habib, and K. Heitmann,
invited book chapter, Publisher: Wiley, submitted.
-
Cosmic Calibration: Constraints from the Matter Power Spectrum and
the Cosmic Microwave Background,
S. Habib, K. Heitmann, D. Higdon, C. Nakhleh, and B. Williams,
Phys. Rev. D76, 083503 (2007),
astro-ph/0702348
|
|