GitLab

ADerrorsCF.jl:

• Added method ADerrors.cov_sym(). It computes the covariant matrix summing the gamma function of the off-diagonal component from -iw to iw. Considering the observables A and B, without assuming detail balance of the simulation algorithm, compute the "positive montecarlo-time gamma function \Gamma_{AB}(t)" as usual. while the "negative montecarlo time gamma function" is computed using the relation, true by construction, \Gamma_{AB}(-t} = \Gamma_{BA}(t). This method is equivalent to ADerrors.cov if detail balance is present and in the limit of infinite statistic. In test/test_cov_sym.jl, this statement is tested for three syntetic data for which we generate a Montecarlo chain of lenght 10 000 000. The covariance matrices computed with the two method differs for ~1% or less

ADerrorsTools.jl:

• Fix to ADerrors.cobs that was in developer branch, was merged into master

ADerrorsUtils.jl:

• Added methods ADerrors.chiexp() that accept a keyword argument C::Array{Float64,2}. If given, then the chi square expected is computed by applying the formula directly, otherwise it uses the method already present in ADerror. This method is more accurate in determining the chi square expected and should be preferred whenever the covariant matrix is already available.
• Updated fit_error() to accept an optional parametere C (default to nothing). If given, and if chi_exp is set to true, C is used to compute the chisquare expected.