Matskási István (szerk.): A Magyar Természettudományi Múzeum évkönyve 94. (Budapest 2002)

Radoinova, D., Tenekedjiev, K. ; Yordanov, Y.: Stature estimation according to bone length in Hungarian population

RESULTS The described methods are realized in an original computer system written in MATLAB 4.2 c.l and described in (TENEKEDJIEV & RADOINOVA 2001). In all models we apply GILES and BORCAN age correction method. The figures in parentheses in formulae (l)-(9) are standard errors of the regression parameters. The results are derived using the following statistical parameters: - significance level for rejecting outliers 0.5% (with a classification of a given observation as an outlier the probability for a error is not more than 1/200) with a maximum of 2 loops (for detecting one level of measurement errors and one level of atypicalness); - significance level for checking heteroskedasticity is 5%; significance level for the ANOVA-test with all models is 5%; - minimal adjusted coefficient of multiple determination R

2

in = 15% for accepting a model eliminating the heteroskedasticity (a model explaining less than 15% of the observed dispersion of the residuals' module is practically insignificant even if statistically significant); - significance level of the t-test for the regression coefficients and of the ANOVA test checking the model adequacy is 5%; - the confidence level of the standard error range is 95%; - the confidence levels of the predicted stature range and of the regression coefficient's regions are 95%, 99%, and 99.9%. I. Hungarian females 1.1. Maximal stature regression on humerus - Three outliers are rejected from the sample - 2 in the first loop and 1 in the second loop (n = 83-3 = 80). Homoskedasticity is rejected in 8 tests: WHITE test (linear) with P-value = 0.0498 and R

2 = 0.036; three GLEJSER tests (linear, reciprocal and squared) with P-value > 0.0143 and 0.055, as well as in the four BREUSCH and PAGAN tests with P-value>0.0215 and 0.036. All of the four models of the residuals' modules are adequate (maximal P-value = 0.026) but neither is acceptable - (maximal R

2

= 0.067

<0.15 = R

2

in

). This proved practically negligible heteroskedasticity. The regression formula for determining the mathematical expectation E[.] of the maximal stature in [ cm] by the length of the humerus (H) in [ cm] is: (1) E[stature] = 89.05 4- 2.252*H (1.701) (0.052)