Number of Mixtures



There has been some suggestion that there may be more than 2 subpopulations. In particular, Wilcox has suggested that 3 populations may better describe the data with one population encompassing the majority of births in the normal range and 2 other populations at each tail, accounting for the extra high and low birth weights compared to a single normal distribution. Our method does not appear to support this arrangement of subpopulations. When we use the mixture method described here with three populations, we get a similar arrangement to the 2 population result with a small “compromised” population and 2 large populations with similar means and standard deviations. We do not get a third population with a very high mean value.