Day 907

Thursday.


1506.05251
Primordial power spectrum features and $f_{NL}$ constraints
Gariazzo, Lopez-Honorez, Mena

The simplest models of inflation predict small non-gaussianities and a featureless PS.  However, there exist a large number of well-motivated theoretical scenarios in which large non-gaussianities could be generated.  In general, in these scenarios the primordial PS will deviate from its standard power law shape.  Study, in a model-independent manner, the constraints from future large scale structure surveys on the local non-gaussianity parameter f_NL when the standard power law assumption for the primordial PS is relaxed.  If the analyses are restricted to the large scale-dependent bias induced in the linear matter power spectrum by non-gaussianities, the errors on the fNL parameter could be increased by 60% when exploiting data from the future DESI survey, if dealing with only one possible DM tracer.  In the same context, a nontrivial bias |delta f_NL|~2.5 could be induced if future data are fitted to the wrong primordial PS.  Combining all the possible DESI objects slightly ameliorates the problem, as the forecasted errors on f_NL would be degraded by 40% when relaxing the assumptions concerning the primordial PS shape.  Also the shift on the non-gaussianity parameter is reduced in this case, |delta f_NL|~1.6.  The addition of CMB priors ensure robust future f_NL bounds, as the forecasted errors obtained including these measurements are almost independent on the primordial PS features, and |delta f_NL|~0.2, close to the standard single-field slow-roll paradigm prediction.


1506.05356
Non-Gaussian forecasts of weak lensing with and without priors
Sellentin, Schäfer

Assuming a Euclid-like WL data set, compare different methods of dealing with its inherent parameter degeneracies.  Including priors into a data analysis can mask the information content of a given data set alone.  However, since the information content of a data set is usually estimated with the Fisher matrix, priors are added in order to enforce an approximately Gaussian likelihood.  Here, compare priories forecasts to more conventional forecasts that use priors.  Find strongly non-G likelihoods for 2d-WL if no priors are used, which is approximated with the DALI-expansion.  Without priors, the Fisher matrix of the 2d-WL likelihood includes unphysical values of Omega_m and h, since it does not capture the shape of the likelihood well.  The Cramer-Rao inequality then does not need to apply.  Find that DALI and MCMC predict the presence of a DE with high significance, whereas a Fisher forecast of the same data set also allows decelerated expansion.  Also find that a 2d-WL analysis provides a sharp lower limit on the Hubble constant of h>0.4, even if the equation of state of DE is jointly constrained by the data.  This is not predicted by the Fisher matrix and usually amused in other works by a sharp prior on h.  Additionally, find that DALI estimates FoM in the presence of non-G better than the Fisher matrix.  Additionally demonstrate how DALI allows switching to a Hamiltonian MC sampling of a highly curved likelihood with acceptance rates of ~0.5, an effective covering of the parameter space, and numerically effectively costless leapfrog steps.  This shows how quick forecasts can be upgraded to accurate forecasts whenever needed.  Results were gained with the public code from github/DALI.