RESULTS 



Sample size: The variance analysis of variances showed significant differences between 

 sample sizes, diameters, and plots for the attack density and gallery length variables, but 

 showed no significant differences between sample sizes for brood. The Student -Newm an -Keul's 

 Multiple Range Test (Steel and Torrie 1960)-"- showed that all sample sizes differed significantly 

 from each other for gallery length, and the 1/10 sq. ft. sample differed significantly from all 

 others for attack density. The variance between diameter for both attack density and gallery 

 length did not show like patterns of occurrence. The test failed to reveal differences between 

 sample heights for attack density, and only top and mid-sample heists differed significantly for 

 brood density. 



The brood density variation was greater than either the attack density or gallery length 

 variation. Thus, inferences drawn from the brood density variable will be used to determine 

 sample size for the sampling plan . 



The Teton data showed downward trend of the coefficient of variation with increase of 

 sample size, but the Wasatch data did not. This distribution of the coefficient of variation is 

 probably due to the effect of zero counts --the larger the sample size the fewer zero counts, and 

 the greater likelihood of reducing variance . 



Sample location: The variances of all three variables were generally greater at breast 

 height and mid-height than at the top sampling height. The only significant difference was be- 

 tween mid- and top heights for the brood density variable. On a percentile basis, the three 

 variables were also more consistent. Variance tends to be greatest at breast height, but since 

 the mean values also tend to be larger at this position, the coefficients of variation are not 

 correspondingly high. 



Aspect: A significant interaction between aspects by plots indicated that there was never 

 complete consistency. In some interactions, as with aspects, the data cannot be treated as if 

 they fit the same distribution. Random placement of samples, by aspect, can eliminate this 

 effect. 



Tables 2,3, and 4 show the results of the variance analysis of variances. 



Tree diameter: Not enough trees were sampled to permit any analysis of brood density 

 in relation to d.b.h. 



Regression analysis showed that both attack density and gallery length were significantly 

 related to tree size. Attack density and gallery length increased with tree diameter (tables 5, 

 6, and 7). 



Size of sample needed: The number of samples (trees) needed for a 20-percent SME at 

 the 2/3 probability level was computed for each sample size and all three variables at d.b.h. 

 The north and south samples were combined because they were not random with respect to each 

 other and in effect constituted a single sample . 



Steel, G. D., and J. H. Torrie. Principles and procedures of statistics. New York: 

 McGraw-Hill Book Co. , Inc. 481 pp. 1960. 



8 



