Fitting Correlated Hadron Mass Spectrum Data

We discuss fitting hadronic Green functions versus time $t$ to extract mass values in quenched lattice QCD. These data are themselves strongly correlated in $t$. With only a limited number of data samples, the method of minimising correlated $\chi^2$ is unreliable. We explore several methods of modelling the correlations among the data set by a few parameters which then give a stable and sensible fit even if the data sample is small. In particular these models give a reliable estimate of the goodness of fit.