Kaszab Zoltán (szerk.): A Magyar Természettudományi Múzeum évkönyve 76. (Budapest 1984)

Demeter, A. ; Lázár, P.: Morphometric analysis of field mice Apodemus: character selection for routine identification (Mammalia)

Variance Fig. 7. Relationship between the entropy and the variance of the characters Character selection by dispersion criterion Using Orlóci's method, a single variable, Dim3, was found to account for about threefourth of the total variance (Table 4), the other variables which made up for 95% of the total variance mainly included external and "minor-axis" skull measurements, each with very low relative importance. It is interesting to compare the cumulative relative importances of the variables and the principal components and factors (Fig. 9). The first 22 variables accounted for 95% of the total variance; similar number of principal components made up for 96.5% and the cumulative curves are very much alike, with the P.C. being consistently higher by 1.5-2.0%. The factors obtained by the principal factor analysis are quite different, the total variance explained only slightly increased after factor VII or VIII and the first 22 made up for only 83% of the total variance. Identification The above analyses were carried out with the aim of reducing dimensionality, but for identification of individual specimens a series of discriminant analyses were carried out (Table 5). Direct and stepwise procedures were used for the subset of variables given in the tables. The greatest degree of separation, and therefore correct classification, was achieved by including all 65 variables (Fig. 10). Stepwise entry of these variables resulted in 13 being selected. When the 35 characters possibly available in skull fragments were analysed, the