1.2.4 Analyse the data

For the analysis of Magnitude Estimation data it is important that the responses of the different informants who rated the stimuli on individual scales are normalised. There are various ways of normalisation, e.g. by expressing the individual results as a percentage and thus making them comparable or by transforming the respective scales to z-scores (cf. Featherston 2004, 2005). The latter “effectively unifies the different scales that the individual subjects adopted for themselves, and allows to inspect the results visually” (Featherston 2004 cited in Hoffmann, Chapter 5).

Of course, there are professional software programs that calculate and (in part) analyse the data obtained. Most researchers use repeated measures of ANOVA (= Analysis of Variance) (cf. Crawley 2005; Field 2003b; Gravetter and Wallnau 1992), which provides two such repeated measurements for each experiment, one testing whether the effects are significant by-subject (F1) and a second checking whether they are significant by-item (F2). Lately it has been suggested, however, that so-called “mixed linear models are statistically superior to repeated measures of ANOVA for the analysis of such data” (Hoffmann, Chapter 5). In the following example study, data was analysed via simple calculations carried out by Microsoft Excel.

Created with the Personal Edition of HelpNDoc: Free HTML Help documentation generator