Number of large trees 



On sales of 1000 to 3000 trees, from 10 to 20 percent of the sample 

 size usually can be devoted to a complete tally of large trees. In 

 ponderosa pine a volume of about 2000 board feet, represented by a 

 tree of around 34 to 36 inches d.boh., was found to be a good divid- 

 ing point in the scale books studied „ In Douglas-fir and larch, a 

 vclurne of 1000 board feet or 1500 board feet usually was used as the 

 lower limit of large trees, depending on the number of such trees in 

 the sale. As sale size increases, the relative size of the sample 

 decreases (see Figure 1), but the num-ber of large trees remains about 

 a constant proportion of total trees. For sales larger than about 

 four thousand trees, large trees can be sampled separately rather 

 than tallied completely, the appropriate sampling rate being deter- 

 mined by the estimated variation and number of trees in the large 

 tree stratumo The variation between strata, that is, betvjeen large 

 and small trees, is thereby eliminated and increased efficiency 

 gained . 



Stratifying into three groups 



A still further reduction in sample size can b^^^ made by dividing the 

 tree volumes into three strata. (Appendix VII.) This was demonstrated 

 on tvjo sales where all large trees vi/ere tallied completely — the 

 most efficient dividing point having been established by trial — and 

 the remaining trees divided into "small" and "medium" strata. The 

 size of the samples needed from each of these two strata vjas computed. 

 These samples, together with the large trees, gave a much smaller 

 total number than was needed in the original unstratified sample to 

 meet the same sampling accuracyo Table 2 lists these results and shows 

 how stratified sampling increases efficiencyo As will be shown later, 

 stratified sampling can be accomplished without undue burden to field 

 men. Inasmuch as variation can be materially reduced by stratifying 

 into three groups, Figure 1 shows sample sizes for coefficients of 

 variation as low as 30 percent. 



Stratifying by defect 



Samples can also be stratified by defect classes if it is knovm that 

 two or more easily distinguished defect conditions exist on the sale. 

 If, however, defect is related directly to volume, as is often the 

 case, then volume strata would cover both forms of variation. 



-13- 



