In this research, most people use a private car as a means of transport over a wide living area. Thus, rather than use the nearest neighborhood library, it is relatively easy to visit a large library that is some distance away. For this reason, we use a binary logit model to estimate variables that have a strong correlation with the selection of a large library. For estimation, we use the results of a subscriber survey at Ishikari Library. The people live in Ishikari City and the Kita Ward and the Teine Ward of Sapporo City.
The logit model has two choices: 1) Mainly use a large library (Ishikari Library) and 2) Use the nearest library. Table 2 shows the explanatory variables: 1) Purpose of library use, 2) Whether a person visits the library accompanied by family, 3) Whether a private car is readily available, 4) Profession of respondent, 5) Assessment of close proximity to home, 6) Time and distance to library from home, and 7) Whether the nearest library is small or medium size. The model estimate is based on data from 134 respondents, excluding those with missing values.
Table 2 shows the maximum likelihood estimates. The log likelihood of the reduced model, which only includes constant terms, is -91.06, the log likelihood of the full model is -71.66, and the likelihood ratio test statistic is 38.79 (Seven degrees of freedom). Because all explanatory variables have none zero coefficients, the null hypothesis is rejected. With McFaddenfs p2 equal to 0.213, model fitness is good; and with the choice probability set at a threshold value of 0.5, the percentage of correct responses works out to 76.1%. From this it is also possible to conclude that the modelfs estimates generally fit actual selection results.
Next is an examination of the coefficient estimates. First, there is a significant correlation between going to the library accompanied by family and using a large library. This usage pattern of driving together in a private car to a large library some distance away to borrow a book is also verified by the individual library selection model. On the other hand, the explanatory variable for individual mobility (Using private car to drive oneself to the library) lacks descriptive force. This suggests the need for a usage behavior model for the household unit that can clarify household mobility.
Second, the significance of positive constant terms suggests that respondents have an unconditional propensity to select a large library. On the other hand, there is no significant change in selection between a large library and either the nearest small or medium size library. This shows that in terms of the attractiveness of a large library, there is no essential difference in scale between a facility with 30,000 volumes and one with 80,000. Third, in a conventional facility selection model, the distance variable is regarded as a dominant factor. In this research, time and distance are not significant. Thatfs because the definition of travel time includes the means of transport to a library. As a result, the difference in distance between the nearest library and a large library is not a factor with most people who in this research use a private car. The significance of gassessment of close proximity to homeh and gaccompanied by familyh suggests that selection is based less on distance and more on how the family should spend the holiday together.
Table 1 Definition of Variables
Use Ishikari Library
Use the nearest library
Main or secondary purpose is as family attendant
|Accompanied by family||1=
Use with family
Use by oneself or with a friend
Have own car or share family car
Don't have family car or canft drive
|Assessment of close proximity to home||(Point)|
|Time, distance||Time, distance to Ishikari Library
Time, distance to the nearest library
The nearest library is at a citizenfs center or district center
The nearest library is a district library
Table 2 Estimation Results
|Coefficient||Standard error||P value|
|Accompanied by family||1.1197||0.4194||0.0076|
|Assessment of close proximity||-0.7760||0.2273||0.0006|
|n= 134 (n0=56/n1=78)
LR= 38.79 to Ô2(7)
Percent Correct Prediction: 76.11%