-
A statistical machine learning approach for benchmarking in the presence of complex contextual factors and peer groups
Authors:
Daniel W. Kennedy,
Jessica Cameron,
Paul P. -Y. Wu,
Kerrie Mengersen
Abstract:
The ability to compare between individuals or organisations fairly is important for the development of robust and meaningful quantitative benchmarks. To make fair comparisons, contextual factors must be taken into account, and comparisons should only be made between similar organisations such as peer groups. Previous benchmarking methods have used linear regression to adjust for contextual factors…
▽ More
The ability to compare between individuals or organisations fairly is important for the development of robust and meaningful quantitative benchmarks. To make fair comparisons, contextual factors must be taken into account, and comparisons should only be made between similar organisations such as peer groups. Previous benchmarking methods have used linear regression to adjust for contextual factors, however linear regression is known to be sub-optimal when nonlinear relationships exist between the comparative measure and covariates. In this paper we propose a random forest model for benchmarking that can adjust for these potential nonlinear relationships, and validate the approach in a case-study of high noise data. We provide new visualisations and numerical summaries of the fitted models and comparative measures to facilitate interpretation by both analysts and non-technical audiences. Comparisons can be made across the cohort or within peer groups, and bootstrapping provides a means of estimating uncertainty in both adjusted measures and rankings. We conclude that random forest models can facilitate fair comparisons between organisations for quantitative measures including in cases on complex contextual factor relationships, and that the models and outputs are readily interpreted by stakeholders.
△ Less
Submitted 16 November, 2020;
originally announced November 2020.
-
Evaluating Tests of Virialization and Substructure Using Galaxy Clusters in the ORELSE Survey
Authors:
N. Rumbaugh,
B. C. Lemaux,
A. R. Tomczak,
L. Shen,
D. Pelliccia,
L. M. Lubin,
D. D. Kocevski,
P. -F. Wu,
R. R. Gal,
S. Mei,
C. D. Fassnacht,
G. K. Squires
Abstract:
We evaluated the effectiveness of different indicators of cluster virialization using 12 large-scale structures in the ORELSE survey spanning from $0.7<z<1.3$. We located diffuse X-ray emission from 16 galaxy clusters using Chandra observations. We studied the properties of these clusters and their members, using Chandra data in conjunction with optical and near-IR imaging and spectroscopy. We mea…
▽ More
We evaluated the effectiveness of different indicators of cluster virialization using 12 large-scale structures in the ORELSE survey spanning from $0.7<z<1.3$. We located diffuse X-ray emission from 16 galaxy clusters using Chandra observations. We studied the properties of these clusters and their members, using Chandra data in conjunction with optical and near-IR imaging and spectroscopy. We measured X-ray luminosities and gas temperatures of each cluster, as well as velocity dispersions of their member galaxies. We compared these results to scaling relations derived from virialized clusters, finding significant offsets of up to 3-4$σ$ for some clusters, which could indicate they are disturbed or still forming. We explored if other properties of the clusters correlated with these offsets by performing a set of tests of virialization and substructure on our sample, including Dressler-Schectman tests, power ratios, analyses of the velocity distributions of galaxy populations, and centroiding differences. For comparison to a wide range of studies, we used two sets of tests: ones that did and did not use spectral energy distribution fitting to obtain rest-frame colours, stellar masses, and photometric redshifts of galaxies. Our results indicated that the difference between the stellar mass or light mean-weighted center and the X-ray center, as well as the projected offset of the most-massive/brightest cluster galaxy from other cluster centroids had the strongest correlations with scaling relation offsets, implying they are the most robust indicators of cluster virialization and can be used for this purpose when X-ray data is insufficiently deep for reliable $L_X$ and $T_X$ measurements.
△ Less
Submitted 31 May, 2018; v1 submitted 1 May, 2018;
originally announced May 2018.
-
X-Ray Emitting Active Galactic Nuclei from z = 0.6-1.3 in the Intermediate and High-Density Environments of the ORELSE Survey
Authors:
N. Rumbaugh,
B. Lemaux,
A. Tomczak,
D. Kocevski,
L. Lubin,
P. -F. Wu,
R. Gal,
L. Shen,
A. Mansheim,
C. Fassnacht,
G. Squires
Abstract:
We studied AGN activity in twelve LSSs in the ORELSE survey, at 0.65<z<1.28, using a combination of Chandra observations, optical and NIR imaging and spectroscopy. We located a total of 61 AGNs across our sample that were successfully matched to optical counterparts in the LSSs. Seeking to study AGN triggering mechanisms, we examined the spatial distribution of the AGNs and their average spectral…
▽ More
We studied AGN activity in twelve LSSs in the ORELSE survey, at 0.65<z<1.28, using a combination of Chandra observations, optical and NIR imaging and spectroscopy. We located a total of 61 AGNs across our sample that were successfully matched to optical counterparts in the LSSs. Seeking to study AGN triggering mechanisms, we examined the spatial distribution of the AGNs and their average spectral properties. We found that AGN populations across our sample had less time since the last starburst than the overall galaxy populations. We did not find any relation between AGN activity and location within the LSSs, suggesting triggering mechanisms which depend on global environment are at most sub-dominant. To focus on differences between our AGNs, we grouped them into four sub-samples based on the spectral properties of their parents LSSs. We found one of the sub-samples, SG0023 & SC1604, stood out from the others: AGNs in this sample were disproportionately luminous, their average time since the last starburst event was the smallest, despite the fact that this was not true of the overall galaxy population in those LSSs, and both the AGNs and the overall galaxy population had the largest fraction of close kinematic pairs, which indicates a higher rate of galaxy mergers and interactions. These results suggest that major mergers are driving AGN activity in SG0023 & SC1604, while other processes are likely triggering less luminous AGNs in the rest of our sample. Additionally, minor mergers are unlikely to play a significant role, since the same conditions that lead to more major mergers should should also lead to more minor mergers, which is not observed in SG0023 & SC1604.
△ Less
Submitted 17 August, 2016;
originally announced August 2016.
-
Incommensurate spin density wave as a signature of spin-orbit coupling
Authors:
Aaron Farrell,
P. -K. Wu,
Y. -J. Kao,
T. Pereg-Barnea
Abstract:
Close to half filling and in the limit of strong interactions the Hubbard model leads a spin Hamiltonian. In its original isotropic form this is the Heisenberg model with antiferromagnetic coupling $J$. On a square lattice there is no frustration and the ground state is a classical antiferromagnet denoted by the commensurate order vector ${\bf Q}=(π,π)$. In this work we show that the inclusion of…
▽ More
Close to half filling and in the limit of strong interactions the Hubbard model leads a spin Hamiltonian. In its original isotropic form this is the Heisenberg model with antiferromagnetic coupling $J$. On a square lattice there is no frustration and the ground state is a classical antiferromagnet denoted by the commensurate order vector ${\bf Q}=(π,π)$. In this work we show that the inclusion of spin orbit coupling (SOC) in the fermionic Hubbard model on a square lattice leads to an incommensurate spin density wave (ISDW) in the strong interactions limit at half filling. In this limit the Hubbard model leads to a non-diagonal and anisotropic spin Hamiltonian which exhibits the incommensurate spin order as its classical ground state. We note that an ISDW can be found in systems regardless of spin orbit-coupling due to nesting and lattice distortion. These effects, however, can usually be recognized by experimental probes.
△ Less
Submitted 17 February, 2014;
originally announced February 2014.
