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

For (4) the ANOVA-test had P-value = 4.10"

5

, R

2 is 30% and R 2 is 26%. The obtained regression model by WLS as described in Material and Method is: (5) E[stature] = 91.35 + 1.210*H + 0.916*T (1.95) (0.29) (0.26) Eleven tests for detecting heteroskedasticity in (5) are performed with 95% confidence level. Two tests detect certain heteroskedasticity (linear and reciprocal GLEJSER test) with P-value < 0.0182. The heteroskedasticity observed is practically negligible as the regular part of the variation of the residuals is maximally eliminated from the model (5). This showed clearly on Fig. 4, where the quantity Gt (squared residuals of WLS model (5) divided by residual variance) is plotted as a function of the predicted stature according to the prescriptions of BREUSCH and PAGAN (MADDALA 1988). The random character of Gt is excellent indicator for the presence of practically negligible heteroskedasticity. The correlation coefficients between the parameters of (5) are as follows: between the intercept and the humerus coefficient: -0.374, between the intercept and the tibia coefficient: 0.183 and between the humerus coefficient and tibia coefficient:-0.980. The estimated standard error is 1.24*e

i

(H,T) cm with a 1.07*ej(H,T) cm to 1.46*e;(H,T) cm 95% confidence interval, where e;(H,T) is the expected reFig. 4. BREUSCH and PAGAN'S quantity Gt (squared residual/residual's variance) as a function of the predicted values of (5) divided to (4)