tree is several times more costly than tree measuring it. The 

 greater the spread in cost betvjeen initial measurement and check 

 scale, the greater should be the accuracy of the first sample in 

 order to reduce the amount of check scaling required to meet the 

 final accuracy standards. The cost ratio can be determined roughly 

 by estimating the time required to do the two jobSo 



Tree measuring costs about 12 cents per treeo Check scaling under 

 favorable conditions costs about 50 cents per tree, based on check- 

 ing 40 trees per day at a cost of ^20, When sample trees have to 

 be searched out, check scaling costs will be higher.. Of course, a 

 portion of the time spent on check scaling can be charged to sale 

 inspection. Often a few trees can be check scaled incidentally to 

 supervision or marking without much added expense. Probably 4 to 1 

 is a good average ratio to use for cost of check scaling to cost 

 of tree measuring until a better one is developed through experience 

 or special studies. 



Figure 2 shows the first sample size required to give a desired 

 accuracy when all factors are considered. This chart should be 

 used for determining sample size vjhenever the correction factor 

 is to be determined by check scaling instead of by a "knowledge 

 correction factor." It gives sample sizes that will estimate log- 

 scaled volume within the desired accuracy after the necessary check 

 scaling is completed and the correction factor applied. The accur- 

 acies specified in Figure 2 are based on a 68 percent probability. 

 (See Appendix VIII. ) 



Figure 3 shows the ratio of check scale sample size to first sample 

 size for various correlation coefficients and cost ratios. The 

 check scale sample size is found by multiplying the number of trees 

 in the first sample, found in Figure 2^ by the ratio shown in Figure 3. 

 It is evident from an inspection of this alignment chart that a high 

 correlation between marker's and checker's volumes reduces the check 

 scaling job. For example, using a correlation coefficient of 0.7 and 

 a cost ratio of four, it is found that over six-tenths of the trees 

 originally measured must be check scaled. If a correlation coeffi- 

 cient of 0.9 could be used, only about one-fourth of the first sample 

 need be check scaled. Thus, by improving efficiency of estimating, 

 the amount of checking is materially reduced. A table in Appendix I 

 gives a rough guide to the appropriate correlation coefficient. 



-21- 



