Accuracy Testing Measures

The Data Set Characteristics

Data consisted of 236 permanent plots from four ongoing experiments: (1) Yellow-Poplar Stand Density Study (Olson 1959, Della-Bianca 1965), (2) Mixed Hardwood Stand Density Study (Beck 1973), (3) Mixed Hardwood Unthinned Control Plots (Beck 1973), and (4) Yields of Unthinned Mixed Hardwood Stands (Smith et al. 1975). The plots were located in North Carolina, Virginia, Georgia, Tennessee, and South Carolina and remeasured at 5 yr intervals. All but 32 plots were thinned from below at the time of installation to obtain a range of basal areas for different site-age combinations. A total of 184 observations were used with time intervals of 10 yr as the minimum and 25 yr as the maximum. The test data set had the following characteristics: even-aged, yellow-poplar or mixed oak dominated overstory, dry-mesic to wet-mesic moisture regimes, 20 to 100 yr in age, and 25 to 230 ft2/ac of basal area. The site index range was 74 to 138 ft (base age 50) for yellow-poplar and 60 to 110 ft (base age 50 for northern red oak, Q. rubra ) for mixed oak (Rauscher et al. 2000).

The Accuracy Testing Performance Measures

In technical parlance, accuracy refers to the size of the deviation between observed and predicted values. Accuracy has two components: bias and precision. Bias refers to the success of estimating the true value of a quantity and precision refers to the clustering of sample values about their own average. The measurements used to assess accuracy were (Rauscher et al. 2000):

• Bias in % : success of estimating the true value of a quantity, the smaller the better.
• Tolerance Interval (TI): a measure of precision, the smaller the better.
• Mean Square Error (MSE): the smaller the better.
• PA-15: the proportion of time the model predictions come within +/- 15% of the observed value; the larger the value the better.

In general, the smaller the bias and/or the greater the precision, the more accurate is the model which translates into a smaller MSE and a larger PA-15 value.