Kinesiology, Vol. 31 No. 1, 1999.
Original scientific paper
Linear and non-linear correlation models of morphologic variables
Franci Ambrožič
; Fakultet za sport, Sveučilište u Ljubljani, Slovenija
Abstract
In this study the adequacy of the linear correlation model was tested on a representative sample of 686 males, 19-27 years of age, by analyzing the relations between twenty-three anthropometric measures, representing four hypothetical morphological factors. The nature of the relations was analyzed using scatter diagrams, hierarchical polynomial regression analysis and Blackman's test of linearity. The analysis of 253 variable pairs gave 91 statistically significant differences between "r" and "ŋ" at a 5% significance level (theoretically there should be 12-13) and 63 at a 1% significance level (theoretically 2-3). Inclusion of quadratic and higher terms in polynomial regression gave a higher explained variance, therefore it appears that the linear model does not describe well the relationship between some of these variables. Further verification, using non-linear factor analysis, seems to be warranted. The presented findings were obtained with a sample of adult males, but they could be as valid for females, children or selected groups of the general population, such as anyone active in sports.
Keywords
morphology, correlation, models, nonlinearity, adults, males
Hrčak ID:
256697
URI
Publication date:
28.4.1999.
Visits: 681 *