The Multidimensional Structure of Physical Fitness: Invariance over Gender and Age
Marsh, Herbert W., Research Quarterly for Exercise and Sport
This investigation extends the factor analytic approach pioneered by Fleishman(1964) by incorporation subsequent developments in the application of confirmatory factor analysis and the physical fitness literature. Specifically, it test the ability of an a priori factor structure of physical fitness to fit (i.e., account for) data base on 25 indicators of fitness (field exercises, technical measures, and laboratory measures) for 2,817 boys and girls ages 9, 12, and 15. An eight-factor (Cardiovascular Endurance, Explosive/Dynamic Strength, Static Strength, Flexibility/Joint Mobility, Blood Pressure, Lung Function, Body Girth, and skinfold) model derived from previous research fit the data well for each of thesix age/gender groups, considered separately. Based on tests of factorial invariance, factor loadings and factor correlations were reasonably invariant across the six groups. This important finding indicates that all 25 indicators are equally valid for boys and girls aged 9, 12, and 15 and that the multidimensional structure of physical fitness generalizes over gender and age.
Key words: physical fitness structure, physical fitness development, confirmatory factor analysis, factorial invariance, gender differences
Physical fitness is a valued goal for men and women of all ages. Of particular relevance to this investigation are continuing concerns about youth fitness. Despite the importance of the physical fitness construct, empirical tests of a priori models of the factor structure of physical fitness and their generalizability over gender and age have not been given adequate attention.
A Construct Validity Approach
Physical fitness is a hypothetical construct; thus its construct validity must be established. In a construct validity approach, investigations can be classified as within-construct studies, which evaluate the internal structure of physical fitness using techniques such as factor analysis, or between-construct studies, which attempt to establish a theoretically consistent, logical pattern of relations between measures of physical fitness and other constructs. The resolution of at least some within-construct issues should be a logicaL prerequisite to between-construct research. This emphasis on construct validity, factor analysis, and within-construct studies of the structure of physical fitness is the focus of the present investigation.
In the physical fitness literature, a distinction is typically made between large sample epidemiology-like studies of youth fitness that rely primarily on easily administered field exercises, which do not require expensive equipment, and small sample laboratory studies of adult (or elite athlete) fitness, which emphasize technically sophisticated measures that require expensive equipment. The distinction, however, has the potential for confusing the indicators of physical fitness with the physical fitness construct and for confusing the technological sophistication required to obtain a measure with the construct validity of a measure. The inexpensively collected field exercises should not be viewed as "poor cousins" of the more expensive laboratory measures, and the technologically sophisticated measures are not necessarily more valid indicators of the physical fitness construct. The purpose of the field exercises is not to provide a necessarily imperfect prediction of the laboratory measures that could be achieved if only adequate resources were available to test all subjects in a laboratory setting. Rather, both the field exercises and the laboratory measures are merely indicators of the physical fitness construct, whose validity should be systematically evaluated within a construct validity approach. Hence there is a need to evaluate the structure of physical fitness in studies that (a) include a wide array of field exercises and sophisticated laboratory measures, (b) are based on sufficiently large samples to appropriately apply statistical techniques such as factor analysis, and (c) test the generalizability of findings over individual characteristics such as age and gender. …