Multivariate Normals




Many researchers suggest that both gestational age and birth weight should be used in research to help account for births that are small for gestational age verses births that are premature. While this does add information, it also adds complication to the model and analysis. We have found that the two-subpopulation multivariate birth weight by gestational age CDDmlr on infant mortality: (a) describes the birth weight by gestational age distribution well, (b) describes the birth weight by gestational age infant mortality surface well, (c) identifies statistically significant heterogeneity in infant mortality between the two subpopulations identified, and (d) typically finds several birth weight by gestational age minimal mortality. The global minimum mortality occurs as expected at birth weights by gestational ages slightly larger than the mean birth weight and gestational age of the population examined and is due to the ''normal'' subpopulation. The local minima usually occur at similar birth weights as the global minimum but at gestational ages about eight weeks shorter and five weeks longer than the global minimum and are due to the ''compromised'' subpopulation. Some of these dynamics may represent misestimation of gestational age. The multivariate birth weight by gestational age CDDmlr model can be used to explore these issues further. In particular the multivariate CDDmlr method unifies the two lines of research; one identifying ''compromised'' births and the other identifying ''erroneous'' gestational age, that have employed mixture models to study birth outcomes.