G. Coenders, B. Ferrer-Rosell
Compositional Data Analysis (CODA) is useful when relative rather than absolute information is predicted or used as predictor, and when absolute information is unavailable. Sometimes, research hypotheses concern both absolute and relative information and both are available. The so-called T-spaces solve the problem of combining relative and absolute information in a statistical model and have already been studied in the dependent role.
This poster shows how to use them in the predictor role: how to construct the variables, how to interpret the results, and advantages with respect to traditional approaches which are typically hard to interpret and lead to high collinearity. With appropriate estimation methods, the approach can be used with dependent metric, binary, ordinal or count variables. An illustration is provided on how trip budget volume and the way the budget is distributed into tourist products and services affects an ordinal measure of tourist satisfaction.
Palabras clave: Compositional data analysis, generalized linear model, T-space, CODA with a total, tourist expenditure
Programado
X04 Pausa Café. Sesión Posters. Reunión TEST - Edificio 1
7 de septiembre de 2016 11:40
Edificio 1
