Attached
The problem of estimating parameters of finite
mixtures, IS one of the oldest estimation problems. Due to
the lack of a completely satisfactory solution, this problem
still attracts a great deal of attention. Other than the
mixtures of normal components, the most widely used
mixture distributions are the mixtures of exponential
components. The simplest is the mixture of two exponential
components whose probability density function is given by,
fex) = pAle -AI
X+ (l-p)A2 e -A2X
=0
for Al' A2> 0 and 0 < p < 1.
x>O
otherwise
Mixtures of this type are frequently applied in life
statistics and failure data. In this thesis, the problem of
estimating parameters of a mixture of two exponential
components is studied.
Our first effort I the use of the method of moments,
did not grve us satisfactory solutions. The simulation study has
shown that the resulting estimates deviated drastically from
the actual parameters, Next, the method of maximum likelihood