ABSTRACT
Panel or longitudinal datasets occur when continuous or discrete observations yit on a set of subjects or units i = 1, . . . , n are repeated over a number of measuring occasions t = 1, . . . , Ti possibly differing between subjects (with N =∑
Ti). There are many contexts for such data to occur, and many variations in study design and data format as well as variations in how subjects or units are defined. For instance in economic and marketing applications (Rossi et al., 2005), the panel is typically at the level of an individual consumer, household or firm, and relates to questions such as economic participation or brand choice (for households or consumers), or investments and outputs at firm level. In actuarial applications (Antonio and Beirlant, 2007; Frees et al., 2001) the “subjects” may consist of groups of policyholders (risk classes) with responses being insurance claim counts. In clinical trials, patients are randomly assigned to treatments and measures taken during follow-up to compare treatment effects.
