Microsoft Word 2007 ichrie conference Proceedings Final-Final 06-06-07. doc

Data Collection 
The data collection method for this study was adapted and modified from Dillman’s (2000) group 
administration of self-administered surveys. An illustrated general protocol for group administration of 
questionnaires includes five steps: introduction, special instructions, distribution, retrieval, and debriefing (Dillman, 
2000). The in-class visit data collection procedure employed in this study was composed of three steps, namely pre-
delivery preparation, an oral survey introduction before distributing packages to potential respondents, and in-class 
distribution, modified from group administration approach. The major draw-back of this data collection method, 
compared with tradition mailing method, is the lack of non-response information. Without their contact information, 
it was impossible to reach potential respondents after the in-class survey delivery. Therefore, this study could not 
generate any information about those who failed to send their surveys to the researcher.
Data Analysis 
The data analysis proceeded in four steps. Firstly, descriptive statistics were conducted, which depicted a 
general picture of mature respondents’ socio-demographic status, and their leisure travel behavior. The SAS 
statements, PROC MEANS, and PROC FREQ, were used to analyze the average rankings, as well as the ranking 
distribution of each item.
Secondly, exploratory factor analysis (EFA) was employed to reveal the possible underlying factor 
structures of travel motivations. The EFA was employed not only to reduce observed constraint variables into a 
smaller number of factors that would account for most of the variance in the observed variables but also to identify 
the underlying factor structures (Hatcher, 1994; Stevens, 2002). A principal component with varimax rotation was 
conducted for variable reduction purposes. The overriding objective was interpretability of the resulting varimax-
rotated factor matrix, not maximum explained variance or the inclusion of all factors with Eigenvalues of 1.0 or 
higher. The desired results met these criteria: 1) the components, after orthogonal rotation, are interpretable; 2) each 
variable loads 0.4 or higher on only one component; and 3) the acceptable reliability coefficient, which was 
determined with Cronbach’s coefficient alpha, should not be lower than .60. PROC FACTOR was used to decide on 
the number of factors retained for each construct.
According to Sheppard (1996), cluster analysis followed by factor analysis was a better approach if the 
primary purpose of a study was segmentation. Therefore, a cluster analysis was conducted as the third step. The 
number of clusters was determined with a two-stage cluster analysis method (Arimond & Elfessi, 2001). The 
statement PROC FASTCLUS was performed to identify the number of clusters. The motivation items retained in the 
factor analysis were independent variables used in the statement PROC CANDISC.

---