Weibull regression in the analysis of the aging phenomenon of biological systems
Georgescu, G.
Revista Medico-Chirurgicala a Societatii de Medici Si Naturalisti Din Iasi 103(1-2): 170-175
1999
ISSN/ISBN: 0048-7848 PMID: 10756906 Document Number: 498327
The processing battery represented by standard regressions is totally insufficient for the primary regression analysis of biomedical observation data. Using only this processing system, it is possible to loose a significant amount of information and especially to obtain false conclusions, seeming to be "statistical justified". This conclusion is sustained by the results obtained in this paper on 8 samples. In this paper it was created a calculation algorithm that allows the automatic, simultaneous effectuation of 16 types of regression, obtained by change of variable. This algorithm finally orders these regressions according to the values of correlation coefficient and emphasizes the regression type that mathematically fits the best the observed data. The choice of the optimal regression type depends basically by the fact that the results supplied by that regression corresponds with the biological characteristics of the studied phenomena. In the specific case of optical nerve, in all the 8 samples and sub-samples analyzed, the regression that supplied the higher correlation coefficients was the Weibull model, with exponent variable as a function of time.