Page 283 - Contributed Paper Session (CPS) - Volume 7

CPS2099 Takatsugu Yoshioka et al.
We can find three centroids that line up along the first axis and cluster 81
component points. We can also interpret the configuration of principal
components and the meaning of variables from the loading vectors such that
the proposed method derives two principal axes dividing 23 items to patients’
attitudes and behaviours.
Next, we examine the performance of RKM with NLPCA using test score
data. We got the data consisting of 40 students’ test score (categorical, five
levels rating) of 9 subjects (we refer this data TSc). Form this data, we
generated a numerical data by assigning a random number to original
categorical score according to the rating score (we refer this data TSo).
Regarding this numerical data as a true continuous structure, we apply RKM
with NLPCA to TSc and ordinary RKM to TSo with k=4 and r=3 and compare
them.
Figure 2 is a biplot of ordinary RKM to TSo and Figure 3 a biplot of RKM
with NLPCA to TSc.

270 | I S I W S C 2 0 1 9