-60- 



Inventory Design 



INVENT analyzes data based on two sampling designs: 



(1) simple random sampling, 



(2) stratified random sampling with simple random sampling within 

 each stratum. 



Sampling Is assumed without replacement — that Is, sample units (plots 

 or points) are far enough apart so that a tree is sampled on only one sample 

 unit. 



INVENT allows for the possibility of double sampling regression estima- 

 tion of volume. The error associated with regression estimation is small 

 relative to the error among sample units. Thus, it is ignored in the 

 discussion given below on sample size approximation. 



Foresters commonly use systematic sampling rather than simple random 

 sampling. Generally, estimates from systematic sampling are somewhat more 

 precise than from simple random sampling provided the systematic sample 

 units are spread well over the tract sampled. Therefore, confidence 

 intervals based on simple random equations provide somewhat conservative 

 estimates for well-representative systematically collected data. For the 

 same reason, sample size is usually overestimated. 



Sample Size 

 Sample size is treated briefly. More detailed explanations are given 

 in Barrett and Nutt (1975) , Cochran (1977) , Freese (1962) , and Dllworth and 

 Bell (1976). 



If a tree is sampled on more than one sample unit, the estimated means are 

 still unbiased but the variance is underestimated. 



