| Abstract: | Biomathematical-methodical aspects are exposed of quantitative recording of mean growth courses which will be taken as basis for comparisons between different sample curves. In the paper the body length growth process of children in the age period from 0 to 3 years is considered. There are samples of children growing up under different social conditions and data of corresponding samples collected about 20 years ago. After remarks on the correlation between physical and psychical development of infants several possibilities for representing mean body length growth curves are discussed. The advantages of empirical regression are explained as the best suited method for a modelfree data evaluation. The nonparametric location trend test of Cox and Stuart is used for statistical proving of global difference or parallelism in the course of 2 mean growth curves. Level differences in the global sense between 2 compared curves may be tested by constructing a Scheffé confidence region for a properly defined constrast. Local level differences can be proved by the t-test for those pairs of corresponding mean values for which the conditions of applying this test are fulfilled. The calculation of a curve-related normal belt as a succession of 95%-tolerance regions (without confidence probability) is demonstrated by an example. |
