A comparison of some methods for estimating mixed normal distributions

Some features of the distribution of birthweight of human infants

“…For both single defined in terms of five parameters, the mean and distribution consists of two standard deviation of each component and the corresponding to the two proportion of one or other component in the mix-[his shows the characteristic ture. In theory, the estimation of these five para-Iand confirms that a repre-meters from grouped data is a straightforward ight distribution in terms of a problem, provided that the grouping is sufficiently d components provides an coarse (see Fryer and Robertson, 1972), although there are certain computational difficulties associated with the need to employ an iterative procedure to determine the optimum value of the object function. In practice, however, the inaccuracy of the basic data and the fact that a very large part of the population belongs to one component of the data was arranged to minimize the effect of irregularities due to digit preference.…”

Mixture Distributions—I

“…, β m is critical. An essentially similar approach is the minimum chi-square * criterion discussed by Fryer and Robertson [18].…”

On the probability distribution of economic growth

“…As noted by e.g. Fryer and Robertson (1972), the method of maximum likelihood breaks down for this model. Unfortunately the choice of MM excludes model selection criteria such as the AIC and BIC.…”