example, if the black marble is the first one drawn, the marker knows 

 that the next nine trees will not be sample trees. Or, conversely, 

 if nine white marbles have been drawn, the marker knovjs that the next 

 tree will have to be the sample treeo 



The "one pocket" system of randomization in which the marble is replaced 

 in the same pocket after each draw reduces the chance of personal bias 

 but does not give assurance that the desired sampling rate will be main- 

 tained. Over a great many draws, if the marbles are carefully mixed each 

 time, the desired ratio would be met, but on a small sale considerable 

 error could be made in the sampling rate and hence in the tree count. 

 Then, vjhen the sample mean is multiplied by the total number of trees 

 to obtain estimated total volume, a product of two errors is involved 

 which greatly reduces the accuracy . Mechanical randomizers have been 

 designed, such as the "Reinhardt Differential Selector," which follow 

 the "one pocket" system, but the disadvantages already mentioned apply 

 to this device as well. 



Where more than one species is involved and the samples are stratified, 

 one man has difficulty in keeping the various sampling rates straight. 

 Crew marking is, therefore, recommended for sample tree measurement., 



THE CHECK SCALE 



Because of the many factors that contribute to errors in tree measure- 

 ment, some of which can be eliminated with experience and some of which 

 must be accepted, check scaling is usually necessary » It is the only 

 way to insure that a tree measurement job is as good as log scaling. 

 If the estimators have measured and recorded tree data accurately and 

 the correlation between their volumes and the scaled volumes is highj 

 the check scaling job is relatively easy. If their vjork was inaccurate, 

 the check scale must be correspondingly larger to insure a proper 

 correction factoro 



Tree size as source of discrepancy 



The correlation coefficient of markers' and checkers' volumes may be 

 low for several reasons » Errors in measuring the trees by the esti- 

 mator, especially in estimating height, often cause discrepancy betv/een 

 tree measured and scaled volumes. Even if no mistakes are madej the 

 common practice of measuring d.b.h. to even inches (two-inch classes) 

 and height to v/hole 16=foot logs introduces a large error. Consider p 

 for example, a ponderosa pine tree recorded as 24 inches In diameter 

 and five logs high, with a volume of 810 board feet. Its diara.eter 

 might actually be anything from 23.0 inches to 24,9 inches, with a 



-16- 



