Censored data implies that a certain inventory variable was somehow not fully measured for all the trees. Assuming truncated values for the taller tress as it is typically observed in forest experiments result in volume under-estimation. Consequently, any inference about the forest pattern may be compromised by using these biased tree height values (truncated values). To overcome this problem, we present a methodology that treats the truncated tree height values differently. Different models were fitted to height-diameter data from black wattle stands. A total of 12 plots were fully inventoried with diameter at 1.30 meters aboveground and total height always being measured. To provide a framework of how dealing with censored data in field experiment analysis, we created a second dataset in which tree heights above 17 m were set to the constant value of 17 m. Hence, it is possible to compare the true regression line to the censored regression. The use of censored regression in the presence of censored data proved to be very powerful, since the tree height values can be substantially improved through estimation instead of being assumed as truncated, which naturally reflects in less biased volume estimation as well.
No datasets are available for this submission.