Digital Signal Processing Reference
In-Depth Information
As a consequence most of the values of the random variable are concentrated
around the mean value
np
, which is a well-known property of normal random
variables. Therefore, in this case, the probabilities (
5.109
) can be approximated
with a normal variable with the parameters:
m ¼ np
s
2
(5.131)
¼ npq:
It follows:
2
e
ð
k
np
Þ
f
X
ðxÞ ¼
X
k¼
0
C
n
p
k
q
nk
dðxkÞ
X
n
n
1
2
pnpq
p
2
npq
dðxkÞ:
(5.132)
k¼
0
Note that the normal density in (
5.132
) approximates the probability mass
function (
5.109
)
In this way, we can use the functions shown in Sect.
4.2.2
to find the approximate
probabilities for a binomial random variable:
Pfkk
1
; ng ¼
X
k
1
1
2
1
þ
erf
k
1
m
C
n
p
k
q
nk
s
2
p
:
(5.133)
k¼
0
Similarly,
Pfk
2
kk
3
; ng ¼
X
k
3
1
2
erf
k
3
m
s
2
p
erf
k
2
m
C
n
p
k
q
nk
s
2
p
:
(5.134)
k¼k
2
Example 5.6.3
In
n ¼
60 independent trials, a binary event
A
occurs with a
probability of
p ¼
0.3. Find the probability that event
A
occurs greater than or
equal to 20 and less than or equal to 40 times.
Solution
The number of occurrences of event
A
is a binomial variable with a mean
value and variance of:
m ¼ np ¼
60 0
:
3
¼
18
:
(5.135)
s
2
¼ npq ¼
60 0
:
3
ð
1
0
:
3
Þ ¼
12
:
6
:
(5.136)
The corresponding normal density approximation is given as:
f
X
ðxÞ
X
k¼
0
n
e
ðk
18
Þ
2
1
2
p
12
p
6
dðxkÞ:
(5.137)
2
12
:
:
6
From (
5.134
), we get:
Pf
20
k
40
;
60
g
"
#
1
2
40
18
2
12
20
18
2
12
p
p
erf
erf
¼
0
:
7134
:
(5.138)
:
6
:
6
Search WWH ::
Custom Search