Mathematical Statistics Anna Janicka
Lecture V, 22.03.2021
PROPERTIES OF ESTIMATORS, PART I
Plan for today
1. Maximum likelihood estimation examples – cont.
2. Basic estimator properties:
◼ estimator bias
◼ unbiased estimators
3. Measures of quality: comparing estimators
◼ mean square error
◼ incomparable estimators
◼ minimum-variance unbiased estimator
MLE – Example 1.
Quality control, cont. We maximize
or equivalently maximize
i.e. solve solution:
x n x
x x n
X P
L − −
=
=
= ( ) (1 )
)
(
) 1
ln(
) (
) ln(
ln )
) 1
ln((
) ln(
ln )
( + + − −
=
− +
+
= − x n x
x n x
l n x n x
1 0 )
(
' =
−
− −
=
x n x
l
n MLE
( ) = ˆ
ML=
xMLE – Example 3.
Normal model: X1, X2, ..., Xn are a sample from N(
,
2).
,
unknown.we solve
we get:
( ) ( )
( )
(
2 2)
2 1 2
2 2
1 2
1
2 ln
) 2 ln(
) (
exp ln
) , (
2 2
n x
x n
x l
i i
n
i n
+
−
−
−
−
=
−
−
=
=
−
=
= +
−
+
−
=
0
0 )
2 (
2 2
3
1
2 1 2
n i
l
i i
l n
x n x
x
1 2
2
( )
ˆ
ˆ
ML=
X,
ML=
n
xi−
X
Estimator properties
Aren’t the errors too large? Do we estimate what we want?
is supposed to approximate
.In general: is to approximate g(
). What do we want? Small error. But:
◼ errors are random variables (data are RV)
→ we can only control the expected value
◼ the error depends on the unknown .
→ we can’t do anything about it...
ˆ
) ˆ X( g
Estimator bias
If is an estimator of
:bias of the estimator is equal to
If is an estimator of g(
):bias of the estimator is equal to
/ is unbiased, if
other notations, e.g.:
) ˆ X(
) ( )
ˆ ( ))
( )
ˆ ( ( )
( E
g X g E
g X g
b = − = −
) =
( ˆ ( ) − ) =
ˆ ( ) −
(
E X E Xb
) ˆ X( g
) ˆ X( ) g
ˆ X(
b
( ) = 0 for
ˆ) (g B
The normal model: reminder
Normal model: X1, X2, ..., Xn are a sample from distribution N(
,
2).
,
unknown.Theorem. In the normal model, and S2 are independent random variables, such that
In particular:
X
) ,
(
~ N
2 nX
) 1 (
~
21 2
2
−
−
S n
n
) 1 2 (
Var and
, , 2 4
2 2
, = = −
S n S
E
X n X
E, = ,and Var, = 2
Estimator bias – Example 1
In a normal model:
= X
ˆ
1
ˆ
1= X
ˆ
2= 5
Estimator bias – Example 1
In a normal model:
is an unbiased estimator of
: is an unbiased estimator of
: is biased:
bias:
= X
ˆ
X = E X = E = X = n =
E
n ni i
n
1 1
1 ,
, ,
ˆ ( )
1
ˆ
1= X
E
, ˆ
1( X ) = E
,X
1=
ˆ
2= 5
2 for
eg
5 5
)
ˆ
2(
,,
=
= =
X E
E
) = 5 − (
b
Estimator bias – Example 1
In a normal model:
is an unbiased estimator of
: is an unbiased estimator of
: is biased:
bias:
= X
ˆ
X = E X = E = X = n =
E
n ni i
n
1 1
1 ,
, ,
ˆ ( )
1
ˆ
1= X
E
, ˆ
1( X ) = E
,X
1=
ˆ
2= 5
2 for
eg
5 5
)
ˆ
2(
,,
=
= =
X E
E
) = 5 − (
b
any model with unknown mean :
Estimator bias – Example 1 cont.
is a biased
estimator of
2: is an unbiased
estimator of
2 :
=−
=
ni i
n
X X
S
11 2
2
( )
ˆ
(
2 2 2)
2 21
2 2
, 1
1 1 2
, 2
,
2 2 )
( )
(
) (
) (
) ˆ (
−
= +
− +
=
−
=
−
=
=
n n n
n i n
i i
n
n n
X n X
E X
X E
X S
E
=−
−
=
ni i
n
X X
S
12 1
2 1
) (
(
2 2 2)
11(
2)
21 1
2 2
1 , 1 1
2 1
1 ,
2 ,
) 1 (
) (
) (
) (
) (
) (
2
=
−
= +
− +
=
−
=
−
=
−
−
= −
−
n n
n
X n X
E X
X E
X S
E
n n n
n i n
i i
n
Estimator bias – Example 1 cont.
is a biased
estimator of
2: is an unbiased
estimator of
2 :
=−
=
ni i
n
X X
S
11 2
2
( )
ˆ
(
2 2 2)
2 21
2 2
, 1
1 1 2
, 2
,
2 2 )
( )
(
) (
) (
) ˆ (
−
= +
− +
=
−
=
−
=
=
n n n
n i n
i i
n
n n
X n X
E X
X E
X S
E
=−
−
=
ni i
n
X X
S
12 1
2 1
) (
(
2 2 2)
11(
2)
21 1
2 2
1 , 1 1
2 1
1 ,
2 ,
) 1 (
) (
) (
) (
) (
) (
2
=
−
= +
− +
=
−
=
−
=
−
−
= −
−
n n
n
X n X
E X
X E
X S
E
n n n
n i n
i i
n
not necessarily the normal model!
Estimator bias – Example 1 cont. (2)
Bias of estimator is equal to
for n → , bias tends to 0, so this estimator is also OK for large samples
=−
=
ni i
n
X X
S
11 2
2
( )
ˆ
b n
2
)
( = −
for any distribution with a variance
Asymptotic unbiased estimator
An estimator of g(
) is asymptotically unbiased, if0 )
( lim
: =
→ b
n
) ˆ X( g
How to compare estimators?
We want to minimize the error of the estimator; the estimator which makes smaller mistakes is better.
The error may be either + or -, so usually we look at the square of the error (the
mean difference between the estimator and the estimated value)
Mean Square Error
If is an estimator of
:Mean Square Error of estimator is the function
If is an estimator of g(
):MSE of estimator is the function
We will only consider the MSE. Other measures are also possible (eg with absolute value)
) ˆ X(
))
2( )
ˆ ( ( ˆ )
,
( g E
g X g
MSE = −
)
2) ˆ (
( ˆ )
,
( =
E
X−
MSE
) ˆ X( g
) ˆ X(
) ˆ X( g
Properties of the MSE
We have:
For unbiased estimators, the MSE is equal to the variance of the estimator
ˆ ) ( Var )
( ˆ )
,
( g b
2g
MSE = +
MSE – Example 1
X1, X2, ..., Xn are a sample from a distribution with mean
, and variance
2.
,
unknown. MSE of (unbiased):
MSE of (unbiased):
MSE of (biased):
= X
ˆ
X n Var
X E
X MSE
2 ,
2
,
( )
) ,
,
( =
− =
=
1
ˆ
1= X
ˆ
2= 5
2 1
, 2
1 ,
1
) ( )
, ,
( X = E
X − = Var
X =
MSE
2 2
,
( 5 ) ( 5 )
) 5 , ,
( = E
− = −
MSE
MSE – Example 2 Normal model
MSE of
MSE of
= n= −
i i
n
X X
S
11 2
2
( )
ˆ
=−
−
=
ni i
n
X X
S
12 1
2 1
) (
1 ) 2
( )
, , (
4 2
, 2
2 2
, 2
= −
=
−
= E S Var S n
S
MSE
4 2
4 2
2 2
4
2 ,
2 2
2 2
, 2
1 2
1 2
) 1 (
) ˆ (
ˆ ) ( ˆ )
, , (
n n n
n n n
S Var
b S
E S
MSE
= −
− + −
=
+
=
−
=
ˆ ) , , ( )
, ,
( S2 MSE S2
MSE
MSE – Example 2 Normal model
MSE of
MSE of
= n= −
i i
n
X X
S
11 2
2
( )
ˆ
=−
−
=
ni i
n
X X
S
12 1
2 1
) (
1 ) 2
( )
, , (
4 2
, 2
2 2
, 2
= −
=
−
= E S Var S n
S
MSE
4 2
4 2
2 2
4
2 ,
2 2
2 2
, 2
1 2
1 2
) 1 (
) ˆ (
ˆ ) ( ˆ )
, , (
n n n
n n n
S Var
b S
E S
MSE
= −
− + −
=
+
=
−
=
ˆ ) , , ( )
, ,
( S2 MSE S2
MSE
in any model: similarly, just with different expressions
MSE and bias – Example 2.
Poisson Model: X1, X2, ..., Xn are a sample from a Poisson distribution with unknown parameter
.ML
= ... = X
ˆ
0 )
( = b
X n X
X
MSE n
i i
n
= =
=1 =
Var 1
Var )
, (
Comparing estimators
is better than (dominates) , if
and
an estimator will be better than a different estimator only if its plot of the MSE never lies above the MSE plot of the other estimator; if the plots intersect, estimators are
incomparable
) ˆ1(X
g gˆ2( X )
ˆ ) , ( ˆ )
, (
MSE g
1MSE g
2
ˆ ) , ( ˆ )
, (
MSE g
1MSE g
2
MSE – Example 1 again
X1, X2, ..., Xn are a sample from a distribution with mean
, and variance
2.
,
unknown. (unbiased)
(unbiased)
(biased)
S2 (biased)
(unbiased)
= X
ˆ
1
ˆ
1= X
ˆ
2= 5
ˆ 2
S
Comparing estimators – Example 1 cont.
We have
From among
is better (for n>1)
are incomparable,
just like
From among is better
1
ˆ1
ˆ = X and
= X
ˆˆ 5
ˆ = and
2 =
Xˆ 5
ˆ1 = 1 and
2 =
X2 2 and Sˆ S
ˆ 2
S
Comparing estimators – cont.
A lot of estimators are incomparable →
comparing any old thing is pointless; we need to constrain the class of estimators If we compare two unbiased estimators,
the one with the smaller variance will be better
Minimum-variance unbiased estimator
We constrain comparisons to the class of unbiased estimators. In this class, one can usually find the best estimator:
g*(X) is a minimum-variance unbiased estimator (MVUE) for g(
), if◼ g*(X) is an unbiased estimator of g(),
◼ for any unbiased estimator we have for
) ˆ X( g
) ˆ(
) (
* X Var g X g
Var
How can we check if the estimator has a minimum variance?
In general, it is not possible to freely minimize the variance of unbiased
estimators – for many statistical models there exists a limit of variance
minimization. It depends on the
distribution and on the sample size.