Digital Signal Processing Reference
In-Depth Information
where
M
is the set of most repeated values, and r is the number of times a member of
M
is repeated in the sample.
Proof:
Following the steps of the proof for the unweighted version, it is
straightforward that
N
ˆ
2
β
=
arg min
X
j
∈M
W
i
(
X
i
−
X
j
)
.
(5.34)
0
=
1
,X
i
=
i
X
j
Dividing by
i
=
1
W
i
, and applying square root to the expression to be mini-
mized, the desired result is obtained.
PROPERTY 7 (Shift and Sign Invariance)
Let Z
i
=
+
X
i
b. Then, for any K and
W
,
(i)
ˆ
ˆ
β
(
Z
1
,
...
,Z
N
)
=
β
(
X
1
,
...
,X
N
)
+
b;
K
K
(ii)
ˆ
ˆ
β
(
−
Z
1
,
...
,
−
Z
N
)
=−
β
(
Z
1
,
...
,Z
N
)
.
K
K
Proof:
Follows from the definition of the weighted myriad in Equation 5.26.
PROPERTY 8 (Unbiasedness)
Let X
1
,X
2
,
,X
N
be all independent and symmetrically distributed around the point
of symmetry c. Then,
ˆ
...
β
K
is also symmetrically distributed around c. In particular, if
E
ˆ
K
exists, then E
ˆ
β
β
=
c.
K
Proof:
If
X
i
is symmetric about
c
, then 2
c
−
X
i
has the same distribution as
X
i
.
It follows that
ˆ
has the same distribution as
ˆ
β
K
(
X
1
,X
2
,
...
,X
N
)
β
K
(
2
c
−
X
1
,
2
c
−
ˆ
X
2
,
...
,
2
c
−
X
N
)
, which from Property 7, is identical to 2
c
−
β
K
(
X
1
,X
2
,
...
,X
N
)
.
It follows that
ˆ
β
K
(
X
1
,X
2
,
...
,X
N
)
is symmetrically distributed about
c
.
5.5.1
Geometrical Interpretation
Weighted myriads as defined in Equation 5.27 can be interpreted in a more
intuitive manner. Allow a vertical bar to run horizontally through the real
line as depicted in Figure 5.9a. Then the sample myriad,
ˆ
β
K
, indicates the
position of the bar for which the product of distances from point
A
to the
sample points
X
1
,X
2
,
,X
N
is minimum. If weights are introduced, each
sample point
X
i
is assigned a different point
A
i
in the bar, as illustrated in
Figure 5.9b.
The geometrical interpretation of the myriad is intuitively insightful. When
K
approaches 0, it gives a conceptually simple pictorial demonstration of the
mode-myriad formula in Equation 5.13.
...
Search WWH ::
Custom Search