Estimation of mixtures of distributions
Eric Järpe
Seminar paper 1996:4, Göteborg University, Göteborg

Abstract
Two problems came up due to an econometrics investigation. They concern maximum likelihood estimation of parameters in a mixed distribution. The stochastic model used, to approach this, is the classical one, simpliest of mixtures W = U X₁ + (1-U) X₂ where X₁ and X₂ are normal random variables (e.g. salaries) of two groups (e.g. half- and full-time employees) and U is a 0-1-variable. The questions is, if possible, how to estimate the different parameters in the probability density function of W.


PDF of a mixture

Email me and i will send you a copy of the whole work (in Swedish).