“…Recently, semi-parametric (Umbach and Wilcox, 1996;Wilcox and Russell, 1983) and fully parametric (Fryer et al, 1984;Gage, 2000Gage, , 2002bGage, , 2003Gage and Therriault, 1998;Oja et al, 1991;Platt et al, 2001;Tentoni et al, 2004) mixture models have been applied to the study of birth outcome distributions. One line of research divides a birth cohort into several probabilistically defined latent subpopulations and interprets them as ''normal'' versus ''compromised'' births (Fryer et al, 1984;Gage, 2000Gage, , 2002bGage, , 2003Gage and Therriault, 1998;Umbach and Wilcox, 1996;Wilcox and Russell, 1983).…”