RELATIVE BIAS
Lognormal Estimator for Gamma Data
otal ee OE
FS
’
a
2
=
ive ~
2?
?
Fi 8loe et
¢
ro
Ome eb @
’
»
¢
6
‘Ss
vo
v
+
-_
ce
?
e
+
v
=1,5% .7 6%
”
.
x
wi
S
x
2
hO
>
OF
8 alsee er
vo
510
@
WwW
»
*
nO
Sample Size n
COVERAGE
Lognormal Estimator for Gamma Data
pen) eH &
°
é
¢
é
ec
2
Sie e+ ee
Sisal vt oe
@
?
Ciel ee ee
oa
Ealeeee
¢
¢
S aleeee
@
¢
-_
culte ee
60
wb
NH
¢
eo
¢
wo
Sample Size n
Fig.
6.
Monte Carlo simulation results on bias and
confidence interval coverage of the lognormal estimator with gamma distributed
data,
EX = 1.
Values at the intersections
of rows and columns are either 100[Ave(E(x))
- EX]/EX (upper figure) or the proportion
of 1000 replications where the computed
confidence interval included the true expected value (lower figure).
622