mean expressed as a percentage of the mean gives the sampling accu- 

 racy. Increasing the number of observations in the sample reduces 

 the size of the standard error and enhances the accuracy of the 

 sampling. Furthermore, by setting the accuracy at some predetermined 

 amount, the size of the sample necessary to meet it can be detenninedo 

 (See Appendix IV.) The process of estimating volume by a sample and 

 evaluating its accuracy was followed in analyzing the data collected 

 from scale books. Since each of these samples was drawn from a scale 

 book in which the actual total volumes were known, the actual accuracy 

 of sampling could be- compared vjith the theoretical accuracy computed 

 from the coefficient 'of variation and the sample sizeo 



CHECK SCALE ANALYSIS 



In analyzing check scale data, the coefficient of variation as well 

 as the correlation, or amount of agreement, between the check scaled 

 and tree measured volumes, are needed. (See Appendix V.) If the 

 correlation coefficient is high (approaching one for perfect agreement) 

 a very small amount of check scaling gives a reliable correction 

 factor. However, if estimated and check scaled volumes of the same 

 trees are poorly correlated, many measurements are necessari/" to deter- 

 mine the true correction factor » The number of trees required in the 

 check scale depends mainly upon the amount of variation in the volumes 

 and the correlation between estimated and scaled volumes of the same 

 trees. Other factors affecting size of the check scale are the number 

 of trees in the first sample and the relative cost of measuring a 

 standing tree and check scaling a treeo 



The amount of check scaling required to meet a given accuracy was 

 determined for the various sales studiedo (See Appendix IXo) 



RESULTS OF SALiPLE TREE MEASUREtlENT STUDY 



RANDOMIZED SAMPLING 



Coefficients of variation of tree volumes vjere found to be rather uniform 

 throughout all the samples» Douglas-fir and larch volumes, hov^evers 

 had slightly lower coefficients of variation than ponderosa pine^ 

 This is to be expected, because the diameter range of ponderosa pine 

 on the average timber sale is greater than that of the other species^ 

 Within a single species, about the same extremes in variation generally 

 vjere found. 



Ponderosa pine timber generally has coefficients of variation of around 

 70 to 80 percent, and Douglas-fir coefficients of around 50 to 60 

 percent. Occasional stands, hovjever, vath scattered, unusually large 

 trees, might have larger variations. Methods will be discussed whereby 

 some of this range in variation can be eliminated in choosing the sample „ 



I 



-9- 



