Scale Development Incorporating Cluster Heat Maps: An Abstract

Exploratory factor analysis (EFA) has long been used to identify factors for construct development. EFA mainly relies on factor loadings, average variance extracted, and item-to-total correlations to build constructs. Unidimensionality, or the existence of a single trait, is the main issue for new scale development. However, the evaluation of unidimensionality is not intuitive when the number of items increases. In addition, no validation has been provided to confirm the dimensions of the factors. As the number of items increases, visualizing different factors to further reveal hidden relationship among items and factors gives researchers a better way to develop constructs. By employing cluster heat maps, this study is one of the first to systematically combine this method with EFA.

Traditional factor analysis relies on factor loadings, average variance extracted, and item-to-total correlations to build constructs. Unidimensionality, or the existence of a single trait, is the main issue for new scale development (Gerbing & Anderson, 1988). However, the evaluation of unidimensionality is not intuitive when the number of items increases. In addition, no validation has been provided to confirm the dimensions of the factors. Using the heat-mapping technique as a supplementary technique for factor analysis, this study contributes to the measurement and dimensionality for scale development. Specifically, performing cluster heat-mapping is beneficial for researchers to identify the dimensionality of the scales. The cluster heat map is unique for representing data points, items, and scales or constructs in a single picture since the picture shows not only both of the horizontal axes for the items and their corresponding dendrogram plot of the hierarchical cluster tree of the items but also two of the vertical axes for all of the data points and their corresponding dendrogram plot of the hierarchical cluster tree of the data points. The change of the color gives a visual representation of the distribution of data points among items and scales. The context of this study is to investigate the issues that are pertinent to marketing researchers, the big data revolution. Primarily, the following variables are identified to be relevant and coherent to help direct research and practice in the big data and marketing research discipline. The variables are big data analytics (BDA), traditional marketing analytics (TMA), knowledge types, data fusion, and knowledge and information fusion.