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)

weights for the individual characters. This method will be abbreviated as SS. — DAVIES (1981) found that the information content (in fact entropy) of the characters was a useful criterion for selecting a subset of 25 from the original 101 characters in a numerical taxonomical study of Hemiptera, therefore we tested the applicability of this approach by calculating the entropy (H) of each of the variables by using the Shannon-Wiener function (CLIFFORD & STEPHENSON 1975). The univariate and multivariate statistical analyses were carried out using programs from two packages: BMDP (DIXON 1981) implemented on the IBM 3031 computer, and SPSS (NIE et al. 1975) implemented on the CDC 3300 computer of the Computer and Automation Institute of the Hungarian Academy of Sciences. Dr. M. Rajczy wrote a FORTRAN program for ORLOCI'S (1973) ranking method. RESULTS Univariate analyses Table 1 contains a selection of univariate statistics. The vertical lines indicate the nonsignificantly different species forming the homogenous subsets. In 36, just over half of the variables there were significant differences between all the three species. When pairs of species were united, differences were not significant between A. microps and A. sylvaticus for 19 variables, whereas only for three variables were there no significant differences between A. sylvaticus and A. flavicollis. For two of the variables, homogenous subsets were created by uniting A. microps with A. sylvaticus in the first step, then in a separate subset, again the medium-sized sylvaticus was found to be non-significantly different from the larger flavicollis. For three variables, all the three species were united in one subset. Without going into any more details of the absolute values of the sample statistics, one conclusion is evident; the discriminating power of at least half of the original variables is questionable when viewed solely from a univariate point of view. However, morphological characters are not independent of one another, so we should not place much faith in these findings. A further illustration of this point is given in Fig. 4, which shows the relative frequency distributions of two selected variables, depicting the actual overlap in the characters that have been found either to be significantly different for all three species, or greatly overlapping for at least one pair of species. Fig. 5 shows a selection of bivariate scatter plots, chosen to illustrate some commonly used combinations of measurements. When Left and Wzga are plotted against Cbsl, an often-used allometric character, two different kinds of pattern may be seen; in Fig 5a, Lefi, another measurement along the major axis of the skull, is relatively greater in A. sylvaticus than in A. microps, whereas in Fig. 5c (Wzga, taken along the minor axis of the skull), the points are situated more or less along a straight line. In the other two diagrams we illustrate scattergrams between variables recommended for owl pellets (Fig. 5b, d). In the plot of Csdi against Lefi, A. microps and A. flavicollis are fairly well segregated, but A. sylvaticus shows great overlap with both, with its cluster of points lying mainly below the imaginary line joining the centroids of microps with flavicollis. In the case of Lefi versus Dim3 the stituation is reversed in that here the cloud of points of sylvaticus lie above the imaginary line. The great overlap between the clusters representing specimens of the three species shows that the pairs of characters recommended by TVRTKOVIC & DZUKIC (1977) and TVRTKOVIC (1979) are of limited utility for identification of the three species. Entropy Table 2 contains the entropy values of the characters. The number of character states of the standardized data (rounded to reflect the original 0.05 mm precision of measurement), was the total number of discrete values. The greatest entropy is possessed by measurements