Abstract

Fuzzy data modelling is very appealing, notably when data is obtained from linguistic questionnaires. In this context, methods of fuzzy statistics analyses are needed. In recent years, we have developed a few fuzzy inference methods to test the mean of a distribution. We can mention the fuzzy anova approach called FANOVA, a test based on fuzzy bootstrapped distribution and a procedure using fuzzy confidence intervals. Proper decision rules have been formulated. Using data from the Health, Ageing and Retirement survey in Europe (SHARE Data), we show how to model the data in a fuzzy sense. Then, we delve into the concepts of the fuzzy inference methods and comment on their implementation. Finally, we proceed to test the mean of the distribution and treat the special case of testing the equality of means of different groups. The results are compared, highlighting the advantages and limitations of the inference procedures. Two particular variables are considered: Childhood health status and Self-perceived health.