arXiv subjects

Jason L. Loeppky

Publications and source records attributed to Jason L. Loeppky.

3 recordsLinked to original sources

arXiv2019

TurbuStat: Turbulence Statistics in Python

We present TurbuStat (v1.0): a Python package for computing turbulence statistics in spectral-line data cubes. TurbuStat includes implementations of fourteen methods for recovering turbulent properties from observational data. Additional features of the software include: distance metrics for comparing two data sets; a segmented linear model for fitting lines with a break-point; a two-dimensional elliptical power-law model; multi-core fast-fourier-transform support; a suite for producing simulated observations of fractional Brownian Motion fields, including two-dimensional images and optically-thin HI data cubes; and functions for creating realistic world coordinate system information for synthetic observations. This paper summarizes the TurbuStat package and provides representative examples using several different methods. TurbuStat is an open-source package and we welcome community feedback and contributions.

astro-ph.IM↗

arXiv2017

Identifying Tools for Comparing Simulations and Observations of Spectral-line Data Cubes

We present a statistical framework to compare spectral-line data cubes of molecular clouds and use the framework to perform an analysis of various statistical tools developed from methods proposed in the literature. We test whether our methods are sensitive to changes in the underlying physical properties of the clouds or whether their behaviour is governed by random fluctuations. We perform a set of 32 self-gravitating magnetohydrodynamic simulations that test all combinations of five physical parameters -- Mach number, plasma parameter, virial parameter, driving scales, and solenoidal driving fraction -- each of which can be set to a low or high value. We create mock observational data sets of ${\rm ^{13}CO}$(1-0) emission from each simulation. We compare these mock data to a those generated from a set of baseline simulations using pseudo-distance metrics based on 18 different statistical techniques that have previously been used to study molecular clouds. We analyze these results using methods from the statistical field of experimental design and find that several of the statistics can reliably track changes in the underlying physics. Our analysis shows that the interactions between parameters are often among the most significant effects. A small fraction of statistics are also sensitive to changes in magnetic field properties. We use this framework to compare the set of simulations to observations of three nearby star-forming regions: NGC 1333, Oph A, and IC 348. We find that no one simulation agrees significantly better with the observations, although there is evidence that the high Mach number simulations are more consistent with the observations.

astro-ph.GA↗

arXiv2016

Monte Carlo based Designs for Constrained Domains

Space filling designs are central to studying complex systems in various areas of science. They are used for obtaining an overall understanding of the behaviour of the response over the input space, model construction and uncertainty quantification. In many applications a set of constraints are imposed over the inputs that result in a non-rectangular and sometimes non-convex input space. Many of the existing design construction techniques in the literature rely on a set of candidate points on the target space. Generating a sample on highly constrained regions can be a challenging task. We propose a sampling algorithm based on sequential Monte Carlo that is specifically designed to sample uniformly over constrained regions. In addition, a review of Monte Carlo based design algorithms is provided and the performance of the sampling algorithm as well as selected design methodology is illustrated via examples.

stat.ME↗