The appendix of McLachlan and Basford (1988) contains the listing of
FORTRAN programs for the fitting by maximum likelihood of normal mixture models
under a variety of experimental conditions.
Jones and McLachlan (1990) have
provided a FORTRAN subroutine
MGT (available at StatLib) for the (maximum likelihood) fitting of
mixtures of univariate normal (log normal) components to data that are grouped
into intervals and may also be truncated.
McLachlan, McLaren, and Matthews
(1995) have provided the FORTRAN subroutine NMFIT that fits a mixture of
g normal (lognormal) components for g=1 and g=2 in
turn to data in interval and truncated form, without need for the user to
specify the starting value for the vector of unknown parameters.
Moreover, if so required, it carries out the likelihood ratio test of
g=1 versus g=2 components by adopting a resampling
approach to the assessment of the null distribution of the likelihood ratio
statistic.