The follomng steps outline the method of selecting the best dividing 

 point of large trees: 



1. From cruise data or from a preliminary sample, compile a frequency- 

 distribution of tree volumes « 



2o Compute the mean, standard deviation, and coefficient of variation 

 from the frequency table » (See appendices II and IIIo) The sample size 

 needed (n) if no stratifying is to be done can be determined from these 

 data using the formula in Appendix 17, or Figure 1. 



3«. Select an arbitrary dividing point for large trees and "remainder." 

 The number of large trees (Nj_) plus the number of remainder (Ng) equals 

 total sale size (N) » 



4. Compute a new mean, standard deviation, and coefficient of variation 

 for the "remainder" group » (Since the large group will be tallied 

 completely, it vjill not contribute to the sampling error.) 



5. Compute the adjusted accuracy (Eg) for this remainder group using 

 formula in Appendix VI. 



6. Using Eg, CVgj and Ng compute a new sample size (ng) . 



7o Add ng and N-|_ to give the total number of trees to be measured. 

 This total will be smaller than n unless the dividing point v/as too 

 low and placed too many trees into the large group. However, it 

 probably will not be the smallest possible number unless a lucky choice 

 v^as made in selecting the dividing point. 



8. Repeat the process with different dividing points until the smallest 

 total (ng-f is obtainedo 



VII. FURTHER STRATIFICATION 



If it is desired to stratify further by dividing the "remainder" group 

 into two strata, the follovjing formulae will give the sample sizes 

 needed. Symbols: 



Number trees Standard Number trees 



Tree size in sale Mean volume deviation in sample 



Small Ng -yi" 



Medium 1m ^ 



Small plus 

 medium N„ 



2 •'2 sg ng 



Large 



Total 



N 



-45- 



