Information Technology Reference
In-Depth Information
Fig. 2.9 Two dimensional visualization of the dissimilarity of the individual views derived
from the reanalysis of the data using the procedure proposed in chapter 3. Views connected
through lines belong to the same cluster. Abbreviations used: Doc=Documentation Expert,
CD=Concept Developer, UE=Usability Expert, PM=Product Manager, SOFTW=Software
Expert, DES=Visual Interface Designer, MKT=Market Expert, U=User
MDS configuration, and b) the ratio of the maximum range of the predicted scores
for an attribute divided by the standard deviation
σ k of the estimation error in the
attribute scores.
In the remaining of this section, we employ these criteria for comparing the three
different analysis procedures: a) the traditional averaging analysis, b) the analysis
proposed in this chapter, and c) the second analysis procedure proposed in chapter
3 that aims at optimizing the goodness of fit for the majority of the attributes.
Table 2.4 depicts the number (and percentage) of attributes being adequately
modeled by the resulting views of the three different analysis procedures. The av-
eraging analysis resulted in a total of 65 or 43 % of all attributes being adequately
modeled by a single two-dimensional configuration. Surprisingly, the analysis pro-
cedure proposed in this chapter, despite resulting in five diverse views, performed
worse than the averaging analysis when we are concerned about the number of at-
tributes being adequately modeled by the resulting views. Even when analyzing
together the attributes of individuals that agreed on the overall dissimilarity of the
stimuli, a wealth of attributes still remained inadequately modeled by the resulting
shared configuration. This challenges the validity of two assumptions made in this
 
Search WWH ::




Custom Search