We present a CMB large-scale polarization dataset obtained by combining WMAP Ka, Q and V with Planck 70 GHz maps. We employ the legacy frequency maps released by the WMAP and Planck collaborations and perform our own Galactic foreground mitigation technique, which relies on Planck 353 GHz for polarized dust and on Planck 30 GHz and WMAP K for polarized synchrotron. We derive a single, optimally-noise-weighted, low-residual-foreground map and the accompanying noise covariance matrix. These are shown, through $\chi^2$ analysis, to be robust over an ample collection of Galactic masks. We use this dataset, along with the Planck legacy Commander temperature solution, to build a pixel-based low-resolution CMB likelihood package, whose robustness we test extensively with the aid of simulations, finding excellent consistency. Using this likelihood package alone, we constrain the optical depth to reionazation $\tau=0.069^{+0.011}_{-0.012}$ at $68\%$ C.L., on 54\% of the sky. Adding the Planck high-$\ell$ temperature and polarization legacy likelihood, the Planck lensing likelihood and BAO observations we find $\tau=0.0714_{-0.0096}^{+0.0087}$ in a full $\Lambda$CDM exploration. The latter bounds are slightly less constraining than those obtained employing \Planck\ HFI CMB data for large angle polarization, that only include EE correlations. Our bounds are based on a largely independent dataset that does include also TE correlations. They are generally well compatible with Planck HFI preferring slightly higher values of $\tau$. We make the low-resolution Planck and WMAP joint dataset publicly available along with the accompanying likelihood code.

The lack of power at large angular scales in the CMB temperature anisotropy pattern is a feature known to depend on the size of the Galactic mask. Not only the large scale anisotropy power in the CMB is lower than the best-fit $\Lambda$CDM model predicts, but most of the power seems to be localised close to the Galactic plane, making high-Galactic latitude regions more anomalous. We assess how likely the latter behaviour is in a $\Lambda$CDM model by extracting simulations from the {\it Planck} 2018 fiducial model. By comparing the former to {\it Planck} data in different Galactic masks, we reproduce the anomaly found in previous works, at a statistical significance of $\sim 3 \, \sigma$. This result suggests the existence of a bizzarre correlation between the particular orientation of the Galaxy and the lack of power anomaly. To test this hypothesis, we perform random rotations of the {\it Planck} 2018 data and compare these to similarly rotated $\Lambda$CDM realisations. We find that, among all possible rotations, the lower-tail probability of the observed high-Galactic latitude data variance is still low at the level of $2.8 \, \sigma$. Furthermore, the lowering trend of the variance when moving from low- to high-Galactic latitude is anomalous in the data at $\sim 3\,\sigma$ when comparing to $\Lambda$CDM rotated realisations. This shows that the lack of power at high Galactic latitude is substantially stable against the "look elsewhere" effect induced by random rotations of the Galaxy orientation. Moreover, this analysis turns out to be substantially stable if we employ, in place of generic $\Lambda$CDM simulations, a specific set whose variance is constrained to reproduce the observed data variance.

Based on recent observations of the cosmic microwave background (CMB), claims of statistical anomalies in the properties of the CMB fluctuations have been made. Although the statistical significance of the anomalies remains only at the $\sim2-3\sigma$ significance level, the fact that there are many different anomalies, several of which support a possible deviation from statistical isotropy, has motivated a search for models that provide a common mechanism to generate them. The goal of this paper is to investigate whether these anomalies could originate from non-Gaussian cosmological models, and to determine what properties these models should have. We present a simple isotropic, non-Gaussian class of toy models that can reproduce six of the most extensively studied anomalies. We compare the presence of anomalies found in simulated maps generated from the toy models and from a standard model with Gaussian fluctuations. We show that the following anomalies, as found in the Planck data, commonly occur in the toy model maps: (1) large-scale hemispherical asymmetry (large-scale dipolar modulation), (2) small-scale hemispherical asymmetry (alignment of the spatial distribution of CMB power over all scales $\ell=[2,1500]$) , (3) a strongly non-Gaussian hot or cold spot, (4) a low power spectrum amplitude for $\ell<30$, including specifically (5) a low quadrupole and an unusual alignment between the quadrupole and the octopole, and (6) parity asymmetry of the lowest multipoles. We note that this class of toy model resembles models of primordial non-Gaussianity characterised by strongly scale-dependent $g_{NL}$-like trispectra.