Table 1 . Proportion of variation explained, , for each of the four models applied 

 to 15 different tables from 1989 Kentucky forest survey 



R2 for Model — 



Attribute 



Rows Columns 



Forest area 



Forest area 



Forest area 



Forest area 



Forest area 



Forest area 



Forest area 



Forest area 



Net cubic-foot 

 volume 



Cubic-foot vol. 

 in Sawlog 



Board-foot 

 volume 



Net cubic-foot 

 volume 



Board-foot 

 volume 



No. live trees 



No. growing- 

 stock trees 



County 

 County 

 County 

 County 

 County 



Forest 

 type 



Owner 

 class 



Forest 

 type 



Owner 

 class 



Forest 

 type 



Stand 

 size 



Site 

 class 



Stocking 



class 



Owner 

 class 



Stocking 

 class 



Stand 

 size 



0.40 0.81 0.89 0.90 0.96 



0.43 0.87 0.93 0.95 0.98 



0.41 0.81 0.87 0.90 0.93 



0.42 0.86 0.92 0.92 0.97 



0.41 0.85 0.92 0.92 0.98 



0.50 0.98 0.98 0.98 0.97 



0.49 0.99 0.99 0.99 0.99 



0.48 0.99 0.97 0.98 0.98 



Species 



Species 



Species 



Tree 

 class 



Species 

 Species 



Species 



Diameter 

 class 



Diameter 

 class 



Diameter 



class 



Species 

 group 



Log 

 grade 



Diameter 

 class 



Diameter 

 class 



0.00 0.76 0.78 0.82 0.84 



0.00 0.77 0.79 0.83 0.83 



0.00 0.72 0.74 0.78 0.84 



0.00 0.84 0.87 0.92 0.88 



0.00 0.66 0.72 0.79 0.81 



0.00 0.49 0.53 0.69 0.88 



0.00 0.49 0.53 0.67 0.87 



Model 5. For the area tables, the estimate for the coefficient 

 b2 ranged from -.5 to -.6, and and were either not 

 significantly different from zero or barely significant at the 

 0.05 level. 



Model 5 had the best overall performance with values of 

 0.93 to 0.99 for the area tables and 0.81 to 0.88 for the 

 remaining tables. However, it performed substantially better 



than Models 3 and 4 only for the two tables involving 

 number of trees per acre. The marginally better performance 

 was attributed to the additional information in the model 

 from the sampling errors of the row and column totals. 



Many tables in forestry are constructed using more than one 

 sampling intensity. This alters the relationship of cell 

 estimate to cell variance, which is the basis for generalized 



3 



