Lattice QCD allows for systematically improvable first principle
predictions that are vital for the phenomenological interpretation of
experimental data in the flavour sector. Obtaining phenomenologically
relevant quantities from lattice data typically requires the extraction
of energies and amplitudes from n-point correlation functions. At a
given statistic, reliable extractions of masses and amplitudes from
Euclidean time correlation functions faces two challenges:
1) the accurate estimation of correlation matrices and
2) control over contamination of excited states that are not taken into
account in the parameterisation of the data.
We present a new and simple-to-use method which addresses both of these
well-known issues by applying a (linear and invertible) filter to the
correlation functions. Since this method is based on a transformation
of the data rather than a regularisation of the correlation matrix, it
introduces no additional systematic errors or modifications to the
interpretation of goodness-of-fit tests. We further discuss how the
filter parameter can be optimised. We stress that for this data-
analysis method no additional correlation functions need to be
generated.
