Table 8 .--Regression equations^ aoeffiaients of determination (R^) ^ and 



standard errors for the key variables 



Variable : 





Regression : 

 equation : 



Coefficient of : 

 determination : 

 (R2) : 



Standard 

 error 



Season- long number of 

 visits 





= 0.51 + 1.26(Xi) 



0.88 



7.02 



Season- long recreation 

 use (hours) 



Ye 



= 6.96 + 1.21(X6) 



0. 79 



118.65 



Season- long number of 

 groups 





= 0.16 + 1.24(X7) 



0.86 



1.97 



where: \i = registration total for individual visits, 



Xg = registration total for expected hours of use, 

 Xy = registration total for group visits. 



Multiple regression analyses were next run to determine whether additional covar- 

 iate information might result in more precise estimates of use, especially for the fall 

 strata. Variables representing one- and two-person groups and local groups were added 

 as covariates. These additional covariates were included because earlier work by Wenger 

 and Gregersen (1964) indicated that small groups and local groups tended to register at 

 lower rates than other groups. Results indicated that although the data for these types 

 of use based on interviews generally improved precision as measured by the standard 

 error of estimate, substitution of registration data for interview data essentially 

 wiped out this improvement. Estimating procedures using interview data don't seem 

 practical because an administrator would ordinarily have to work with registration data. 



Regressions were also run to determine whether numbers of registered users might 

 also give an adequate estimate of hours of use and number of groups. It was found that 

 registered visits could be used as well as registered groups in the prediction of total 

 number of groups, but the registered visits variable was considerably poorer than 

 registered visitor-hours in the prediction of total visitor-hours. 



Mechanical Counters 



Mechanical counters were used in conjunction with interview information to evaluate 

 devices that might be used for a several-year period following calibration to obtain 

 inexpensive and acceptably accurate estimates of use. The study was concerned mainly 

 with evaluating counting devices, rather than attempting to establish prediction equa- 

 tions based on their results. 



29 



