Multivariate, non linear analysis of the preference for preservation of the teeth

Application of classical, linear approaches to analyze data of a categorical or discrete nature might bring about false results, because the data are forced to fit inappropriate models. The present paper describes a non linear, multivariate approach which might be considered a principal component analysis of data collected in a survey. The computer algorithm HOMALS is used to detect which dimensions are to be found in the data. Searched for are factors associated with the variable preference for preservation of the teeth. The analysis enabled the description of one subgroup: those who do not care about retaining their teeth are characterized, in a descending order of importance, by irregular dental attendance, a (self-assessed) very poor oral health condition, being equally afraid of extraction and cavity preparation, dissatisfaction with dental treatment in the past, and lowest level of education. It appeared that the categories of the variables are not linear. It is concluded that the combination of characteristics, describing a subgroup in the sample, does not imply the description of another subgroup with opposite characteristics.