Modeling in an oral health study through two statistical methods in Uruguay
In epidemiological studies it is common practice to work with binary variables that reflect the presence of certain diseases, which in turn may be associated with another set of variables, that in general are assumed as risk factors of the former. In the field of epidemiological studies referred to oral health, it is common to inquire about the relationship between the presence of some pathologies and certain characteristics of the study participants through generalized linear models (GLM).
... r models (GLM). However, this type of analysis is usually carried out for each variable of interest separately and at no time is a measure obtained that summarizes the status of each participant. The objective was to apply and compare two methodologies; one applying classical approach of explaining each oral disease separately from a set of explanatory variables and another using item response theory (IRT) models (specifically the Rasch model) since they allow the joint analysis of a set of variables obtaining an individual assessment as a by-product, which in this case is interpreted as 'sickness proneness'. On the other hand, the analysis presented here extends the Rasch model including a linear predictor that allows to investigate about the possible effect of several factors on the propensity of the individuals to suffer the different pathologies. Our results found evidence of an effect of gender, insufficient physical activity (IPhA) and age on general proneness to oral diseases.