Is salience robust? A heterogeneity analysis of survey ratings

Abstract

Differing weights for salience subdimensions (e.g. visual or structural salience) have been suggested since the early days of salience models in GIScience. Up until now, however, it remains unclear whether weights found in studies are robust across environments, objects and observers. In this study we examine the robustness of a survey-based salience model. Based on ratings of N_{o}=720 objects by N_{p}=250 different participants collected in-situ in two different European cities (Regensburg and Augsburg) we conduct a heterogeneity analysis taking into account environment and sense of direction stratified by gender. We find, first, empirical evidence that our model is invariant across environments, i.e. the strength of the relationships between the subdimensions of salience does not differ significantly. The structural model coefficients found can, hence, be used to calculate values for overall salience across different environments. Second, we provide empirical evidence that invariance of our measurement model is partly not given with respect to both, gender and sense of direction. These compositional invariance problems are a strong indicator for personal aspects playing an important role.