Abstract: The measurement scales of tourist involvement and tourist destination image are adapted from previous researches. Involvement scale is adapted from Gursoy & Gavcar's study on international leisure tourist involvement, which includes three dimensions: pleasure/interest, risk probability, and risk consequence. Destination image scale is adapted from Baloglu & McCleary's study on destination image formation, which includes cognitive image and affective image. The data from nearly 900 tourists visiting Guilin and Yangshuo were collected during October and November 2008, and the response rate was 94.1%. In order to avoid the influence of inbound sample heterogeneity on the comparison between inbound and domestic sample, only Anglo inbound tourists' data (n = 349) and domestic tourists’data (n = 292) are used in this research. In this study, a structural equation model is used for theory testing and development. First, confirmatory factor analysis is used to test the applicability of tourist destination image and involvement scales, as well as to test the total measurement model for all of the samples. Second, Multi-groups factor analysis is used to test measurement equivalence and investigate the structural relation among three dimensions of involvement and two image components (cognitive image and affective image). Then, the path coefficients of two structural models from two different group tourists are analyzed and compared.
The results shows tourist involvement has significant effect on destination image. Specifically, for all of the samples, pleasure/interest dimension has positive significant effect on cognitive image and affective image, risk consequence dimension has negative significant effect on cognitive image and affective image, and risk probability dimension has no influence on affective image. The effects of risk probability on cognitive image are different between inbound and domestic tourists. For domestic tourists, risk probability has significant positive effect on cognitive, but for inbound tourists, no significant effect is found. Between the two samples, all path coefficients are equivalent except two coefficients: pleasure/interest→cognitive image and risk probability→cognitive image.
The conclusions are drawn as follows. Tourist involvement has significant effect on destination image, and there is no significant cross-group difference except the effect of pleasure/ interest and risk probability on cognitive image.

Key words: tourist destination image, tourist involvement, simultaneous structural equation analysis for several groups, Yangsuo