Occam's razor meets WMAP

TL;DR

The paper applies information-theoretic model-selection criteria (AIC, BIC) and a memory-based criterion (HIC) to WMAP data to evaluate three claimed large-angle CMB features: departures from scale invariance, a low quadrupole, and the axis of evil. It finds that all razors discount significant evidence for a low quadrupole and scale-non-invariance, while the axis of evil can appear robust under a simple, low-parameter axis model but remains sensitive to dataset choices and foreground treatments. The work highlights how different penalties yield different conclusions and emphasizes the need for careful modeling of systematics and the potential of polarization data to resolve outstanding ambiguities.

Abstract

Using a variety of quantitative implementations of Occam's razor we examine the low quadrupole, the ``axis of evil'' effect and other detections recently made appealing to the excellent WMAP data. We find that some razors {\it fully} demolish the much lauded claims for departures from scale-invariance. They all reduce to pathetic levels the evidence for a low quadrupole (or any other low $\ell$ cut-off), both in the first and third year WMAP releases. The ``axis of evil'' effect is the only anomaly examined here that survives the humiliations of Occam's razor, and even then in the category of ``strong'' rather than ``decisive'' evidence. Statistical considerations aside, differences between the various renditions of the datasets remain worrying.