We describe an approximate statistical model for the sample variance distribution of the non-linear matter power spectrum that can be calibrated from limited numbers of simulations. Our model retains the common assumption of a multivariate Normal distribution for the power spectrum band powers, but takes full account of the (parameter dependent) power spectrum covariance. The model is calibrated using an extension of the framework in Habib et al. (2007) to train Gaussian processes for the power spectrum mean and covariance given a set of simulation runs over a hypercube in parameter space. We demonstrate the performance of this machinery by estimating the parameters of a power-law model for the power spectrum. Within this framework, our calibrated sample variance distribution is robust to errors in the estimated covariance and shows rapid convergence of the posterior parameter constraints with the number of training simulations.

Several cosmological measurements have attained significant levels of maturity and accuracy over the last decade. Continuing this trend, future observations promise measurements of the statistics of the cosmic mass distribution at an accuracy level of one percent out to spatial scales with k~10 h/Mpc and even smaller, entering highly nonlinear regimes of gravitational instability. In order to interpret these observations and extract useful cosmological information from them, such as the equation of state of dark energy, very costly high precision, multi-physics simulations must be performed. We have recently implemented a new statistical framework with the aim of obtaining accurate parameter constraints from combining observations with a limited number of simulations. The key idea is the replacement of the full simulator by a fast emulator with controlled error bounds. In this paper, we provide a detailed description of the methodology and extend the framework to include joint analysis of cosmic microwave background and large scale structure measurements. Our framework is especially well-suited for upcoming large scale structure probes of dark energy such as baryon acoustic oscillations and, especially, weak lensing, where percent level accuracy on nonlinear scales is needed.

The complexity and accuracy of current and future precision cosmology observational campaigns has made it essential to develop an efficient technique for directly combining simulation and observational datasets to determine cosmological and model parameters; a procedure we term calibration. Once a satisfactory calibration of the underlying cosmological model is achieved, independent predictions for new observations become possible. For this procedure to be effective, robust characterization of the uncertainty in the calibration process is highly desirable. In this Letter, we describe a statistical methodology which can achieve both of these goals. An application example based around dark matter structure formation simulations and a synthetic mass power spectrum dataset is used to demonstrate the approach.