ABSTRACT
This chapter provides an overview of the basic statistical approaches for the sensory evaluations of sound. It explains to convey the message that there is insight and value to be gained from going from simple attribute-by-attribute one- and two-sample averaging procedures potentially combined with similarly simple error bars to the more elaborate mixed model analysis procedures. Mixed models are the generic statistical term for models where multiple error/noise components are part of the model. The mixed model approach presented can be viewed as including repeated measures analysis but with focus on correlation and variability structures across and within individuals related to various design factors. The mixed model validation step can be seen as the standard linear model validation with some additional validations of the random effect normality assumptions. The chapter explores attribute-by-attribute univariate analysis and describes visualised univariate results in multivariate ways.
