Section xxx \citep{Davies_2017} presented a Bayesian approach to a significantly zero-censored data set of growth-strain from splitting tests. For more details about growth-strain see \citet{Archer_1987}. At the time, they speculated that i) if significant quantities of tension wood were to develop early during growth (i.e. near the pith) ii) to be later surrounded by normal wood, iii) new outer wood would constrain the shrinkage of the tension wood developed near the pith, creating a growth strain profile where the most tension exists toward the middle-to-centre of the stem, rather than the periphery, as would be the normal profile assumed in literature (presented by \citet{Archer_1987}). As a result, when the splitting test was conducted, the two sides would pull together resulting in the zero values in the dataset. Some strain could not be released as the two halves restrained each other.
In order to test this explanation, a trial of 54 Eucalyptus bosistoana trees was set up, with two treatments and a control (18 individuals in each), planted in an irrigated nursery site at Harewood, Christchurch, New Zealand (the same site as the \citet{Davies2017} study). All individuals were tied to bamboo stakes, the control was staked straight for the two year duration of the experiment. The first 'early' treatment was staked on a 45 degree angle to the vertical and the direction of lean was changed by either 90 or 180 degrees every 4 to 8 weeks depending on the growing season in order to maximise tension-wood production. The angle was chosen as to make sure the tree was lent away from prevailing winds on the site to reduce wind damage. This cycle was continued for the first 12 months, the 'early' treatment trees were staked vertically for the remainder of the experiment. The 'late' treatment trees were staked vertically for the first year of the experiment and bent using the same scheme as the 'early' treatment for the second year. The control produced tension wood as a normal (staked) tree would, where needed as micro-environmental changes resulted in the need for righting its stem, the 'early' treatment (likely) created tension wood near the pith, and normal wood near the periphery, and the 'late' treatment (likely) created normal wood near the pith and tension wood near the periphery. 
Table \ref{130815} shows descriptive statistics from the surviving trees of each treatment, under-bark diameter and growth-strain means with standard deviations (by treatment). Results of an analysis of variance performed in R using MCMCglmm \cite{R,mcmcglmm} for diameter and growth-strain are shown in Table \ref{382816}. The treatments had an effect on growth, the early treatment stunted growth, but little to no effect on growth strain. Figure \ref{353401} shows the boxplots of the treatments on diameter and growth-strain. This experiment shows it is unlikely that tension wood development early in growth was the cause of the zero censored data in the \citet{Davies_2017} study.