SAMPLE DESIGN 



Sampling was concentrated on 

 more heavily used trails. 

 The trail to Glacier Lake, 

 shown here, was by far 

 the most popular. It accounted 

 for close to half of all group 

 visits. It was also the most 

 heavily sampled trail. 



The basic sample design was stratified sampling. This concentrated field time on 

 trails having enough use to enable us to observe the registration behavior of visitors; 

 thus we reduced the number of wasted sample days that didn't yield any data. Trail use 

 was expected to have a highly skewed distribution- -a few trails heavily used and many 

 very lightly used. We also suspected that some types of trails might have different 

 registration patterns at certain times. 



Seven strata were established, differentiated on the basis of varying combinations 

 of three variables: season (summer or fall), time of week (weekday or weekend), and 

 expected use on trails (see table 1). Trails were placed in one of four classes of use 

 on the basis of judgment estimates made by Forest Service Ranger District personnel. 

 The four classes were 1/4 to 1 (very light), 1 to 2 (light), 2 to 5 (medium), or over 

 5 (heavy) visits per day, on the average. Trails estimated to average less than one- 

 fourth of a visit per day were not included (table 1). 



Two-day sampling units, rather than 1-day units, were used to reduce (a) the 

 frequency of zero observations, (b) travel expense, and (c) variability in the estima- 

 tion equations. The number of sample units allocated to each strata was based on the 

 following: (a) The expectation that heavy-use strata would be more variable than light- 

 use strata; (b) that interviewing costs would be much less in the heavy-use strata due 

 to more visitor contacts per workday; and (c) that use estimates were desired for 

 individual trails. We felt that 11 to 18 sampling units in each stratum would produce 

 an adequate regression estimate for each trail even if the corresponding estimates of 

 precision (the estimates of error terms) might be unreliable. Lightly used trails 

 were intentionally sampled at a lower sampling intensity than believed desirable for 

 statistical reasons because more intensive sampling would not have been economically 

 feasible. The numbers of trails and 2-day sample units in each strata are shown in 

 table 1. 



9 



