! Maximum Likelihood Estimation with censored data;
! Keywords: Maximum Likelihood Estimation, Censored Data,
Statistics;
SETS:
obs: stock, leftover;
ENDSETS ! Renard de Lac newspaper data;
DATA:
! Simple [mean,sd] of (stock-leftover)
= [101.46, 13.10];
stock, leftover =
110 17
85 0
110 18
110 15
110 14
110 15
110 10
110 6
100 0
110 10
95 10
95 0
100 20
100 0
110 11
115 17
115 9
115 11
115 17
115 0
115 18
115 15
115 0
135 0
150 13
150 37;
ENDDATA
! Find mu and sigma that maximize the log of the
likelihood that we would observe the above data,
assuming the demands are Normal with mean mu
and standard deviation sigma. Some of the
observations are "censored" in the sense that
we do not know the exact demand, only that demand
was at least equal to the amount of product
we stocked(which we do know);
MAX =
! Cases where demand = sales = stock - leftover;
@SUM(obs(i)| leftover(i) #gt# 0:
-.5* @LOG( 2*3.1415926)- @LOG(sigma)
-.5*((stock(i)- leftover(i)- mu)/sigma)^2)
! Cases where we do not know demand, only
that demand >= sales = stock;
+ @SUM(obs(i)| leftover(i) #eq# 0:
@LOG(1 - @PSN((stock(i)-mu)/sigma)));
|