Statistical analysis tools

Detailed Description

Tools for the statistical analysis of Monte Carlo data and other sampled observables.

The Statistical analysis tools module is a small, header-mostly toolkit for analysing samples produced by Monte Carlo (or any other) simulations. It is built around two ideas: streaming accumulators that update mean, error and autocorrelation estimates on the fly without storing the full sample stream, and standalone functions that operate on a range of pre-collected samples.

Every component has an MPI-aware variant — either an *_mpi overload (e.g. triqs::stat::mean_mpi, triqs::stat::jackknife_mpi) or a member function taking an mpi::communicator (e.g. triqs::stat::lin_binning::mean_error_and_tau, triqs::stat::log_binning::mean_errors_and_taus). The same client code therefore runs serially or in parallel, and per-rank partial results are combined into a single estimator on all ranks.

Topics
	Accumulators
	Streaming binning accumulators that estimate mean, error and autocorrelation time on the fly.
	Concepts
	C++20 concepts that constrain the value and range types accepted by Statistical analysis tools.
	Mean and error estimation
	Free functions that compute the mean / sum and a corresponding error estimate from a range of samples.
	Resampling techniques
	Resampling algorithms for bias-corrected estimates of functions of averages.
	Utilities
	Histograms and small helpers shared by the rest of the Statistical analysis tools module.