Data analysis was conducted in R \cite{Team2015} and RJAGS \cite{Plummer2015}, utilizing a Bayesian approach to estimate the heritability of
growth-strain at the family level. The effect of coppicing is included
as a fixed effect. Specimen groups were grown during different time
periods and for different rotation lengths, group is included as a
random effect. The “Splitting” test is physically constrained to
positive values, as the opening cannot be reduced in the presence of
compression at the stem surface and tension at the pith resulting in
left censored data. Bayesian frameworks provide the ability to simulate
partially observable data, and therefore reducing the systematic errors
which occur due to left censoring. Here the left-censored initial values
are sampled from a uniform distribution between -1.5 and 0.