© 1995 Oxford University Press
research-article |
Etiologic Fraction Analysis for Continuously Distributed Outcome Variables and Empirical Analogy with Dichotomized Outcome Variables
* Department of Epidemiology and Biostatistics Montreal, Quebec, Canada.
Department of Pediatrics, Faculty of Medicine, McGill University Montreal, Quebec, Canada.
Reprint requests to: Dr S W Wen, The Institute for Clinical Evaluative Sciences in Ontario, G-2, 2075 Bayview Avenue, North York, Ontario M4N 3M5, Canada.
Background. It is not clear from the published literature whether R2 estimated from linear regression models for continuosly distributed outcome variables is analogous to the ethiologic fraction for dichotomized outcome variables. This article attempts to address this issue.
Method. Continuous and dichotomous outcomes of the same underlying atttributes (gestational age and fetal growth) were compared using data from a recent study of birthweight distributions in ethnic Caucasian infants.
Results. The relative magnitudes of the etiologic fraction and R2 were quite similar for the same underlying attributes. For example, R2 and etiologic fraction for weight gain rate ranked 2 and 3, respectively, for fetal growth and ranked 4.5 and 5, respectively, for gestational duration.
Conclusions. R2 estimated from linear regression models for continuously distribuled outcome variables appears analogous to the ethologic fraction for dichotomized outcome variables. If due consideration is given to the underlying biological mechanisms of the studied attributes, R2 can be used as a measure of public health impact.
Received 1 September 1994