Open Collections

Statistically calibrating comparison metrics of star formation simulations.

University of British Columbia

We have developed an experimentally designed statistical framework, which allows for observation and simulation position-position-velocity cubes to be compared in a systematic manner. The comparison between cubes can be performed using a multitude of metric measures, in particular we have used the following metrics: principal component analysis (PCA), spectral correlation function (SCF), Cramer, and maximum mean discrepancy (MMD). We ran a suite of carefully chosen simulations, and performed a simulation-to-simulation comparison to test the framework in a controlled environment. We have tested the effects of four physical parameters, Mach number, magnetic field, driving scale, and temperature. The developed framework extracted key physics from the optimally designed experiment. In addition, merits of the different measures were highlighted in a standardized manner. After all four metrics were analyzed, we have concluded that PCA, Cramer and MMD respond in a linear manner to changes in factor settings, where as SCF displayed signs of curvature. Furthermore, all four statistics picked out a consistent physical trend, predicting that Mach num- ber, temperature, and Mach-number-temperature interaction are significant dissimilarity drivers. Finally, we propose that this framework will be used in future work for the calibration of simulations to observational data.

Text

2013-08-01

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/44799

Interdisciplinary Studies

University of British Columbia

UBCO

http://creativecommons.org/licenses/by-nc-nd/4.0/