TESTS OF OVERALL BIAS Sample Clauses

TESTS OF OVERALL BIAS. The range of statistical tests completed between predicted and PSP volumes (of the 432 PSPs at last measurement) included calculations of root mean square error (of prediction), Goodness- of-fit index, bias, and percent bias. The total bias between the final NSYT’s and PSP volumes was determined by using the AVI stand age geo-referenced with each PSP at the last measurement, to look up the predicted volume from the yield group averaged final NSYTs. Tests were calculated from each PSP at last measurement by yield group and species group (Table 24), with weighted average statistics calculated for each cover group (Figure 19) and overall (Table 26), weighted by the net landbase area of each yield group. These tests show that the bias in predicted volumes is generally quite low at less than 10% across the majority of yield groups, specifically when compared across the main species groups of each particular yield group (Table 24). The higher observed percent bias in those secondary species groups with minor representation is primarily due to the low predicted and observed volumes of those species. The exception is for yield group 4 (primarily a BW component), where the deciduous component over-predicted volume when compared to the PSP data by over 20%. The reason for this increased bias is likely due to a combination of factors. The sample size of the PSP data yield group 4 is extremely low (n=6), which has a major impact on resulting statistics. In addition, the GYPSY manual cautioned against modeling pure BW groups because of known modeling difficulties. Given the small area in the THLB (<1%) and that the deciduous operators considered the yield projections reasonable for the deciduous strata, the yield group 4 NSYT provides reasonable representation of the growth of BW stands in the FMA. The G&Y monitoring will identify this as a step to accumulate more information in BW stands for future yield curve development.