Folia Historico-Naturalia Musei Matraensis - A Mátra Múzeum Természetrajzi Közleményei 12. (1987)
Czájlik, P.: A Talpa romana ehiki n. subsp. leírása, koponya méreteinek biometriai elemzése
143 E.CAPANNA (1981) presented the standard deviation, the mean and the standard error of distinctive skull and body measurements for Talpa romána terra typica, S. Italian Talpa romána , Talpa romána stankovici , using PETROV's (1971) data as well. These data were compared with those of the Macedonian and Italian populations of Talpa europaea . To fit the Hungarian Talpa europaea population and Talpa romána ehiki into this set of data, the respective values were calculated^ Cc! f~. Tab . 3) . The table shows , that the Pusztapó population of Talpa romána ehiki n. subsp. corresponds to the other Talpa romána populations. Similarly, the values of the Hungarian Talpa europaea population can be fitted to those of other Talpa europaea populations. Table 3 also shows, that standard deviation (S) and standard error (S ) values for the Pusztapó population (a random sample) are on the same order Ss those of CAPANNA's and PETROV's samples and the Hungarian Talpa europaea sample. This serves the basis for further biométrie comparision of the populations studied. 1.20. Confidence intervals of the means As second step, the confidence intervals of the means of skull measurements in Talpa romána ehiki (Pusztapó) and Talpa europaea (Hungary) were studied at confidence coefficients of 0,95 and 0,99 (c. f . Tables 4, 5, 6, 7, 8). The tables show that the confidence intervals of all distinctive skull measurements do not overlap at a confidence coefficient of 0,99. In other words the two samples were taken from different statistical populations, at a 1 \ level of significance . 1.30. Comparision of means and testing the significance of their differences. To answer the question, whether the samples were taken from the same statistical population, and to value the difference between the two populations, the means were compared, and the significance of their differences were tested. In the second halves of Tables 4, 5, 6 the comparision of the means of rostral width, maximal diameter of maxillar M. and mean length of maxillar set of teeth in talpa romána ehiki and Talpa europaea (Hungary) are presented. The data show, that the two populations are not from the same statistical population. The means of the three skull measurements are defferent at a 0,1 % level of significance . 1.40. Differences in maximal diameter of maxillar M. Since MILLER (1912), SCHWARZ (1948) and CAPANNA (1981) regarded maximal diameter of maxillar M, as distinctive characteristic between Talpa romána populations, the differences in this data were studied in more detail. The résulte show, that Talpa romána ehiki is different from the two Italian Talpa romána populations at 1 4 level of significance, whereas it is not defferent significantly from Talpa romána stankovici . Similarly, the two Italian Talpa romána populations are different from the Macedonian population at 1 % level of significance. The résulte imply, that Talpa romána stankovici and Talpa romána ehiki are clostr relatives with each other, than with the two Italian populations, and vice versa. 2.00. Regression analysis To select the least climate dependent diagnostic specific characteristics, and to study the behaviour of certain skull measurements is Talpa romána and Talpa europaea populations, simple- and multiple linear regression analyses were carried out. Methods : For the calculations Commodore 64 computer and TI-57 TEXAS programmable calculator were used. Simple- and multiple linear regression equations were computed using the MLINREG program. It fits ay=a +a,+...ax curve (here 1< т<^12) to N points using the minimum SO method? The program enaBles us to eliminate a 0 . This makes possible the solution of linear equations with "m" unknowns, the у = ax linear equation is solved by a = (x 7 X.) x matrix equation. Inversion is carried out by using Gauss-Jordan method. To reach the desired precision, the program can also change the rows and columns. If N-m + 2 the program carries out regression analysis. Correlation coefficient (r), r^ , the value of F test and the standard error of curve fitting are calculated.
