IMSE2132

Bill KP Chan
Office: HW8-14
Tel.: 3917 7059
Email: [email protected]
 Point Estimators
• Assume that the population follows a certain
distribution with ! parameters ("! , "" , … , "# ). Our
objective is to estimate the parameters based on a
random sample of size & '! , '" , … , '$ .
• Based on '! , '" , … , '$ , the estimators ("(! , "(" , … , "(# )
are calculated where "(% is an estimate of "% , ) =
1,2, … , !.
• ("! , "" , … , "# ) are unknown constants to be estimated.
("(! , "(" , … , "(# ) are statistics calculated based on the
random sample, i.e., the r.v.s '! , '" , … , '$ .
 Point Estimators
• (&! , &" , … , &# ) are unknown constants to be estimated.
(&*! , &*" , … , &*# ) are statistics calculated based on the r.v.s +! , +" , … , +$ .

Example Time to failure , of 9 independent transistors of the same

type are observed:
620, 710, 1050, 2100, 2510, 2790, 3050, 3320, 3630
Assume that , has an exponential distribution with constant failure rate λ per
unit time. Estimate 5.
*
,~789: 5 ; < = 1, &! = 5, +! , +" , … , +$ = 620, 710, … , 3630 . Find 5.
Example Estimate the parameters > and ? of a Gamma random
variable , if the observations are
4.4, 29.94, 6.26, 17.94, 4.54, 10.05, 8.45, 2.45
,~C>DD> >, ? ; < = 2, &! = >, &" = ?,
+! , +" , … , +$ = 4.4, 29.94, , … , 2.45 . Find >, I
E and ?.
 Two methods
1. Method of moments
2. Method of Maximum Likelihood
 Method of Moments
The method: Equate sample moments to population moments
Sample moment:
1 1 # 1
!! = % &" , !# = % &" , …, !$ = % &"$
$ $ $
If there are ) unknown parameters, set up ) equations.
!! = *(&) Note that
!# = *(& # ) 1 1
! "! = ! & '"! = ! & '"!
% % %
!% = *(& )
1
⋮ = & ! '"! = ! ' !
%
!$ = *(& $ )
Since there are ) equations and ) unknowns (the parameters), the
unknown parameters (moment estimators of the parameters) can be
solved.
Example 1 A continuous r.v. + has pdf J 8 = K + 1 8 % (0 ≤ 8 ≤ 1),
where K > 0 is a constant. The following values of + are observed:
0.2, 0.4, 0.5, 0.7, 0.8, 0.8, 0.9, 0.9
Estimate the unknown parameter K using the method of moments.
How many parameters?
1 parameter. Thus, put !( = . /
Solution
!( = 0.2 + 0.4 + 0.5 + 0.7 + 0.8 + 0.8 + 0.9 + 0.9 /8 = 0.65
( (
. / = 0 12 1 31 = 0 4 + 1 1 *+( 31 = 4 + 1 / 4 + 2
) )
Putting !! = # $ , we have
46 + 1 0.65×2 − 1
0.65 = ⇒ 46 = ≈ 0.857
6
4+2 1 − 0.65
Example 2 Time to failure . of 9 independent transistors of the
same type are observed:
620, 710, 1050, 2100, 2510, 2790, 3050, 3320, 3630
Assume that . has an exponential distribution with constant failure rate
λ per unit time. Estimate 6.
How many parameters?
1 parameter. Thus, put !( = . ;
Solution
. ; = 1/<
620 + 710 + 1050 + 2100 + 2510 + 2790 + 3050 + 3320 + 3630
!( =
9
≈ 2198
Putting !( = . ; , we have
1
<= ≈ ≈ 4.55×10,-
2198
Example 3 Use the observations in the above example
620, 710, 1050, 2100, 2510, 2790, 3050, 3320, 3630
and assume that + has a Rayleigh distribution
/- !
, - = /- 0 exp − if - > 0, , - = 0 if - ≤ 0
2
Estimate /. How many parameters?
1 parameter. Thus, put 9" = : +
Solution
" " $% ! # " " &'
! " = ∫! % & '(%) *% = ∫! +% # exp −
#
*% =
$
& ∫! 1 ! 2 *1

$% ! # ( # ) ) *
where 1 =
#
. Hence, ! " = $
&Γ
#
=
$
& &Γ
# #
=
#$
Also, =) ≈ 2198. Putting =) = ! " , we have
D
C+ = ≈ 3.25×10 &+
2×2198#
Example 4 Use the method of moment to estimate
the parameters f and g " of a normal random variable ,.
The observations are 103.84, 99.26, 99.64, 99.86, 102.25, 98.99.

How many parameters? 2 parameters. Thus, put !( = . ;

and !. = . ; . .
Solution
"
7 , = f, h>< , = 7 , " − 7 , ⇒ 7 , " = f" + g "
103.84 + 99.26 + 99.64 + 99.86 + 102.25 + 98.99
D! = ≈ 100.64
6
103.84" + 99.26" + 99.64" + 99.86" + 102.25" + 98.99"
D" =
6
≈ 10131.59
Thus, putting 7 , = D! and 7 , " = D" , we have
fE = D! = 100.64
fE" + gE " = D" ⇒ gE " = D" − D!" ≈ 10131.59 − 100.64" ≈ 3.18
In general, suppose that observations @(, . . . , @/ are observed
from ;. Then
@( +··· +@/
. ; = !( ⇒ Ĵ = … (1)
L
. .
@( + ⋯ + @ /
. ; . = !. ⇒ Ĵ . + QP . = … (2)
L

Thus, (1) and (2) gives

Ĵ = !(
QP . = !. − !(.
Example 5 Use the method of moment to estimate the parameters
7 and 8 of a Gamma random variable . if the observations are
4.4, 29.94, 6.26, 17.94, 4.54, 10.05, 8.45, 2.45
Given that the mean and variance of a Gamma r.v. are
. ; = ]^ and _]` ; = ]^ . respectively.
How many parameters? 2 parameters. Thus, put !( = . ;
Solution
and !. = . ; . .
#
! 0 = 12 and 314 0 = 12 # = ! 0 # − ! 0
⇒ ! 0 # = 12 # + 1# 2 #

4.4 + 29.94 + 6.26 + 17.94 + 4.54 + 10.05 + 8.45 + 2.45

"$ = ≈ 10.5
8
4.4# + 29.94# + 6.26# + 17.94# + 4.54# + 10.05# + 8.45# + 2.45#
"# = ≈ 184.48
8
Thus, putting ! 0 = "$ and ! 0 # = "# , we have
12 = 10.5 and 12 # + 1# 2 # = 184.48 ⇒ 10.52 + 10.5# = 184.48 ⇒ 2 ≈ 7.06
$%.'
Thus, 1 = ≈ 1.49
(.%)
In general, if the observations are !9 ,· · · , !: ,

$ % = '9 ⇒ *) ,+ = '9 … (1)

;
;
$ % = 1*2 % + $ % = *) ,+ ; + *) ; ,+ ;
$ % ; = '; ⇒ *) ,+ ; + *) ; ,+ ; = '; … (2)

'9;
1 and 2 ⇒ *) =
'; – '9;
'; – '9;
and ,+ =
'9
 Method of Maximum Likelihood
Let ! be a r.v. whose pdf "1 ($) depends on an unknown vector of parameters
&. Let !! , !" , . . . , !$ be a random sample of !. The likelihood function of the
sample space is defined as

)(&) = "1 (!! ) · "1 (!" ) · · · "1 (!$ )

, = -($! , $" ,· · · , $$ ). The

Suppose that )(&) attains its global maximum at &
maximum likelihood estimator MLE of & is defined by:

, = - !! , !" , … , !$
.

To find the global maximum, an effective way is to find the critical points of
the log-likelihood function ln )(&) . Since the log function is monotonic
increasing, if ln )(&) attains its global maximum at &,, )(&) will also attain
,
its global maximum at &.
The steps of the method of maximum likelihood are
1. Identify the pdf 9< (:) (or pmf in discrete case).
2. Construct the likelihood function ;(<) from the
pdf and the sample
3. Calculate the log-likelihood function
4. Find the critical point(s) of ln ;(<) by putting
all first order partial derivatives with respect to
>9 , >; , … , >= of ln ;(<) to zero.
5. Solve for >+9 , >+; , … , >+= from the r equations
resulted in step 4.
Example 1 A continuous r.v. $ has pdf
% & = ' + 1 & ! (0 ≤ & ≤ 1)
where ' > 0 is a constant.
Observed values of $ are: 0.2, 0.4, 0.5, 0.7, 0.8, 0.8, 0.9, 0.9
Estimate '.
Solution The likelihood function > ? =
? + 1 0.2 & B ? + 1 0.4 & B ? + 1 0.5 & B ? + 1 0.7 & B
? + 1 0.8 & B ? + 1 0.8 & B ? + 1 0.9 & B ? + 1 0.9 &
= ? + 1 ' 0.2×0.4×0.5×0.7×0.8×0.8×0.9×0.9 &
' &
≈ ?+1 0.014515
G$ > ? ≈ 8 B ln ? + 1 + ? B ln 0.014515
J 8
G$ > ? =0⇒ + ln 0.014515 = 0
J? ?L + 1
'
⇒ ?L = −1 − ≈ N. OP
() *.*!,-!-
Note: the method of moment estimator #" ≈ 0.857
Example 2 Time to failure . of 9 independent transistors of
the same type are observed:
620, 710, 1050, 2100, 2510, 2790, 3050, 3320, 3630
Assume that . has an exponential distribution with constant failure rate
λ per unit time. Estimate 6.

Solution The pdf of an exponential distribution is

%* / = 0 1 2 +,-
3 0 = 0 1 2 +,./01 1 0 1 2 +,.231 1 0 1 2 +,.3141 1
0 1 2 +,.0311 1 0 1 2 +,.0341 1 0 1 2 +,.0251 1 40 1
2 +,.6141 5 1 0 1 2 +,.6601 1 0 1 2 +,.6/61 = 05 1 2 +35271.,
Thus, ln f < = 9 ln < − 19780<
1
− 19780 = 0 ⇒ ,
2 9
ln 6 3 =0⇒ 2 ≈ 8. 99×;<*@
≈
23 3; 2198
Note: the method of moment estimator 2= ≈ 4.55×10*A
Example 3 Use the observations in the above example
620, 710, 1050, 2100, 2510, 2790, 3050, 3320, 3630
and assume that ; has a Rayleigh distribution
2 @ = <@ A exp(−<@ ./2)
Estimate <.
Solution > 6 = 6. B Q! B R /012* /# where

B$ = 620×710×1050×2100×2510×2790×3050×3320×3630
B# = 620# + 710# + 1050# + 2100# + 2510# + 2790# + 3050# + 3320# + 3630#

Thus, ln ) 2 = 9 ln 2 + ln B! − 2 D B" /2
F 9 B" 18
ln ) 2 =0⇒ − = 0 ⇒ 2= ≈ ≈ H. HH×;<*B
F2 2= 2 B"

Note: the method of moment estimator 2= ≈ 3.25×10*.

Example 4 Use the method of maximum likelihood to estimate
the parameters S and T # of a normal random variable ..
The observations are 103.84, 99.26, 99.64, 99.86, 102.25, 98.99.
! ! )*E "
Solution The pdf is "C J = exp −
"-D " D

! F !
The likelihood function ) N, O " = exp − -(N)
"-D "D!

where - N = 103.84 − N " + 99.26 − N " + 99.64 − N " +

99.86 − N " + 102.25 − N " + 98.99 − N "
!
Thus, ln ) N, O " = −3 ln 2Q − 3 ln O " − -(N)
"D!
G G
ln ) N, O " = 0 and ln ) N, O " = 0
GE GD!
, I E !
⇒ -H N = 0 and − + D %= 0
D! " D
!
⇒N= 103.84+99.26+99.64+99.86+102.25+98.99 ≈ 100.64 and
F
O " = - N /6 ≈ 3.18

Note: the method of moment estimators 4 ≈ 100.64 and : & ≈ 3.18

Example 5 Let / be a binomial r.v. with number of trials ! and
an unknown probability of success g, i.e., /~ijL !, g .
Let /(, /., … , // be a random sample of /. Show that the MLE for
g is ∑/EF( /E /(!L).
Solution The pmf for a binomial r.v. is `C Y = D
E
` E D/E
a where
a = 1 − `. The likelihood function
! ! ! E FE F⋯FE
> ` = … ` , * 0 1 − ` D?/ E, FE* F⋯FE0
Y! Y# Y?
D D D
Thus, ln > ` = ln E, E*
… E0
+ Y! + ⋯ + Y? ln ` + !$ − Y! + ⋯ + Y? ln(1 − `)

J Y! + ⋯ + Y? !$ − Y! + ⋯ + Y?
ln > ` =0⇒ − =0
J` ` 1−`
Y! + ⋯ + Y?
⇒ `̂ =
!$
In general, the likelihood function of a normal random variable . is
+ -. * + -. * + -. *
! / , * ! / * * ! / 0 *
> S, T # = R */ B R */ … R */
#=> #=> #=>
! ? ! ? #
= exp − ∑ Y" − S
#=> #> * "@!
# ? ? ! ?
Thus, ln > S, T = − ln 2Z − ln T # − *
∑ "@! Y" − S
#
# # #>
A A
From ln > S, T # = 0 and ln > S, T # = 0, we get
AB A> *

! !
S[ = ∑?"@! Y" = Y̅ and T[ # = ∑?"@! Y" − Y̅ # = ]?#
? ?

Hence the MLE of S and T # are &^ and _?# .

Note The MLE of T # in this example is _?# ,

#
not the sample variance _?/! .
 Problems of Point Estimation
Example The diameters of the cups produced in a production line are
normally distributed with unknown mean N (cm) and known variance 0.2"
(cm2). Random sample of size 7 from the production line is drawn and the
diameters are measured as follows (unit: cm).
7.89, 7.75, 8.33, 8.04, 7.64, 7.65, 8.23
Estimate the mean N.
! !
Sample mean !X = ∑ $J = 7.89 + ⋯ + 8.23 ≈ 7.93.
$ .
If we use sample mean to estimate the population mean, we say, the
population mean [ ! is approximately equal to !X = 7.93.
Does it mean that the population mean equals 7.93? Probably not!
How confident are you on your estimation?
Very accurate? Quite accurate? Not accurate at all? I do not have a clue!
How to quantify your confidence on the estimation?
Use interval estimation instead of point estimation.
 Example The diameters of the cups produced in a production line
are normally distributed with unknown mean N (cm) and known variance 0.2"
(cm2). Random sample of size 7 from the production line is drawn and the
diameters are measured as follows (unit: cm).
7.89, 7.75, 8.33, 8.04, 7.64, 7.65, 8.23
Find the 95% confidence interval for the mean N.

Definition Let &! , … , &? be a random sample from a random

variable & with parameter d and let dL! and dL# (dL! < dL# ) be two
statistics from the random sample. Suppose given any probability
1 − f, 0 < f < 1 , we have
g dL! < d < dL# = 1 − f
The interval dL! , dL# is called the confidence interval (abbreviated C.I.)
for d with confidence coefficient (or the degree of confidence) 1 − f
(or 100 1 − f %). The end points dL! and dL# are called the lower- and
upper-confidence bounds respectively.
\K and &
Note: & is an unknown constant but & \L are random variables.
 O QP$ < Q < QP# = 1 − S

U8
V U9
V
U8
V U9
V
U8
V U9
V
U8
V U9
V
U8
V U9
V

U8
V U9
V

U8
V U9
V
U8
V U9
V
U8
V U9
V
U8
V U9
V
U8
V U9
V
Q=T
Assume
1. &~j S, T # (Population is normally distributed)
2. &! , … , &? is a random sample of size $ from population &.
3. &^ and _?/!#
are the sample mean and sample variance.

/m and l/,(
.
are statistically independent
/m − J
~ o 0,1
Q/ L

.
L − 1 l/,( . L−1
~ n
Q.

/m − J
~ @(L − 1)
l/,(/ L
Example The diameters of the cups produced in a production line are
normally distributed with unknown mean N (cm) and known variance 0.2"
(cm2). Random sample of size 7 from the production line is drawn and the
diameters are measured as follows (unit: cm).
7.89, 7.75, 8.33, 8.04, 7.64, 7.65, 8.23
Find the 95% confidence interval for the mean N.

!X − N
! = 0.05 ]= ~ ` 0,1
O/ ^
Solution We have O = 0.2, ^ = 7
!X = 7.89 + 7.75 + 8.33 + 8.04 + 7.64 + 7.65 + 8.23 /7 ≈ 7.93
..O,*E
Thus, −Z&.&"N < ] = < ]&.&"N . From a statistical table of the standard
&."/ .
normal distribution, it can be found that the upper 2.5% point of b&.&"N is
;. cd.
95% = O −1.96 < X < 1.96
7.93 − T
−1.96 < < 1.96
0.2/ 7
0.2 0.2
⇒ 7.93 − 1.96× < T < 7.93 + 1.96×
7 7 Area =α/2 Area =α/2
⇒ 7.785 < T < 8.081
-zα/2 zα/2
Exercise A firm produces light bulbs that have a length of life that is
approximately normally distributed, with a standard deviation of 40 hours. If
a random sample of 30 bulbs has an average life of 780 hours,
(a) find a 96% C.I. for the population mean of all bulbs produced by this firm;
(b) how large a sample is needed if we wish to be 96% confident that our
sample mean will be within 10 hours of the true mean?

Solution (a) We have O = 40, ^ = 30, !X = 780.

.Q&*E
Thus, ] = . From a statistical table of the standard normal distribution,
A&/ ,&
it can be found that the upper 2% point of b&.&" is e. <9Hf.
96% = O −2.0537 < X < 2.0537

780 − T
−2.0537 < < 2.0537
40/ 30
40 40
⇒ 780 − 2.0537× < T < 780 + 2.0537×
30 30
⇒ 765 < T < 795
Exercise A firm produces light bulbs that have a length of life that is
approximately normally distributed, with a standard deviation of 40 hours. If
a random sample of 30 bulbs has an average life of 780 hours,
(a) find a 96% C.I. for the population mean of all bulbs produced by this firm;
(b) how large a sample is needed if we wish to be 96% confident that our
sample mean will be within 10 hours of the true mean?

Solution (a) We have O = 40, ^ = 30, !X = 780.

.Q&*E
Thus, ] = . From a statistical table of the standard normal distribution,
A&/ ,&
it can be found that the upper 2% point of b&.&" is e. <9Hf.

(b) We have O = 40, !X − N = 10 and from a statistical table of the standard

normal distribution, the upper 2% point of b&.&" is e. <9Hf.
!X − N !X − N
−2.0537 ≤ O ≤ 2.0537 ⇒ ≤ 2.0537
O/ ^
^
10
⇒ ≤ 2.0537 ⇒ ^ ≈ 67
40/ ^
 Note
• If the sample size is unchanged, it is not
possible to increase the "accuracy" and the
level of confidence of the interval estimations
at the same time.
• The confidence interval for ! when " is known
is Q Q
/m − pZ/. A , /m + pZ/. A
L L
Example The diameters of the cups produced in a production
line are normally distributed with unknown mean S (cm) and unknown
variance. Random sample of size 7 from the production line is drawn
and the diameters are measured as follows (unit: cm).
7.89, 7.75, 8.33, 8.04, 7.64, 7.65, 8.23
Find the 95% confidence interval for the mean S.

Solution We have, % = 7, ! = 0.05

;

& '" = 7.89 + 7.75 + 8.33 + 8.04 + 7.64 + 7.65 + 8.23 ≈ 55.53
":$
55.53
∴ '] = ≈ 7.93
7
;

& '"# = 7.89# + 7.75# + 8.33# + 8.04# + 7.64# + 7.65# + 8.23# ≈ 440.97
":$
; ; ; #
1 55.53#
& '" − '] # = & '"# − & '" ≈ 440.97 − = 0.4585
% 7
":$ ":$ ":$
;
1 0.4585
#
∴ ^;2$ = & '" − '] #
≈ = 0.0764
%−1 6
":$
Notation Let q 0 < q < 1 be a real value number.
@Z L is used to denote the point associated with a r.v.
;~@ L such that,
`
s ; ≥ @Z L =0 2a (1) 31 = q
_! /
That is, @Z L is the point such that the area to its right is q.

Example From a t-distribution table,

@).)( 12 = 2.6810
 /m − J
~ @(L − 1)
l/,(/ L
We have, % = 7, '] ≈ 7.93
#
and ^;2$ ≈ 0.0764
7.93 − T
−b%.%#' 7 − 1 ≤ ≤ b%.%#' (7 − 1)
0.0764/ 7
From a statistical table of the t-distribution, it
Area =α/2 Area =α/2
can be found that b%.%#' 6 = d. eef
-tα/2,n-1 tα/2,n-1

7.93 − N The confidence interval for J

−J&.&"N 6 ≤ ≤ J&.&"N 6
0.0764/ 7 when Q is unknown is
j$*!
7.93 − N h!X − JR/" (^ − 1) D ,
⇒ −2.447 ≤ ≤ 2.447 ^
0.0764/ 7 j$*!
!X + JR/" (^ − 1) D i
⇒ 7.68 ≤ N ≤ 8.19 ^
Example The diameters of the cups produced in a production
line are normally distributed with unknown mean S (cm) and unknown
variance. Random sample of size 7 from the production line is drawn
and the diameters are measured as follows (unit: cm).
7.89, 7.75, 8.33, 8.04, 7.64, 7.65, 8.23
Find the 99% confidence interval for the variance T # .

Solution We have, ^ = 7, ! = 0.01

$
k !J = 7.89 + 7.75 + 8.33 + 8.04 + 7.64 + 7.65 + 8.23 ≈ 55.53
JS!
$
k !J" = 7.89" + 7.75" + 8.33" + 8.04" + 7.64" + 7.65" + 8.23" ≈ 440.97
JS!
$ $ $ "
1 55.53"
k !J − !X " = k !J" − k !J ≈ 440.97 − = 0.4585
^ 7
JS! JS! JS!
$
1 0.4585
"
∴ m$*! = k !J − !X " ≈ = 0.0764
^−1 6
JS!
Notation Let q 0 < q < 1 be a real value number.
nZ. L is used to denote the point associated with a chi-square
r.v. u~n . L such that,
`
s u ≥ nZ. L =0 2j (1) 31 = q
" /
i!
That is, nZ. L is the point such that the area to its right is q.

Example From a n .-distribution table,

. .
n).)g 25 = 37.652 and n).hg 20 = 10.851.
 "
^ − 1 m$*! " ^−1
~ n
O"

#
We have, % = 7, ^;2$ ≈ 0.0764.
# 7 − 1 0.0764 #
∴ k%.<<' 7−1 ≤ ≤ k %.%%' 7−1
l#
From a statistical table of the chi square
distribution, it can be found that Area=α/2 Area=α/2

#
k%.<<' 6 = n. ofpf and
#
k%.%%' 6 = qr. pp χ21-α/2, n-1 χ2α/2, n-1

# 0.4584 #
∴ O k%.<<' 6 ≤ ≤ k%.%%' 6 = 99%
l#

# 0.4584 0.4584 #
0.4584
k%.<<' 6 ≤ ⇒ 0.6757 ≤ ⇒l ≤ ≈ 0.6784
l# l# 0.6757
0.4584 # 0.4584 # ≥
0.4584
≤ k%.%%' 6 ⇒ ≤ 18.55 ⇒ l ≈ 0.0247
l# l# 18.55
∴ 0.0247 ≤ l # ≤ 0.6784
Exercise
The weights of 10 bags of rice are 10.2, 9.7, 10.1, 10.3, 10.1, 9.8, 9.9,
10.4, 10.3, and 9.8 kg. Assuming an approximate normal distribution, (a)
find a 99% C.I. for the mean of all such bags; (b) find a 99% C.I. for T # .
Answer: (a) 9.81, 10.31 (b) (0.023, 0.313)
Solution ^ = 10, o = 1 − 99% = 0.01 [,-.
VU ± X)/+ Y − 1 ×
JR/" ^ − 1 = J&.&&N 9 = 3.25 Y
" " 0.246
n!*R/" ^ − 1 = n&.OON 9 = 1.735 = 10.06 ± 3.25×
10
" " = 9.807,10.313
nR/" ^ − 1 = n&.&&N 9 = 23.59
!&&.F
∑$JS! !J ≈ 100.6, !X = = 10.06 +
!& Y − 1 [,-.
∑$JS! !J" ≈ 1012.58 At d = 0.01, ~g + Y − 1
e+
! Hence, we have
∑$JS! !J − !X " = ∑$JS! !J" − ∑$JS! !J "
$ 9×0.0604
!&&.F! +
g/.112 9 ≤ +
≤ g/.//2 9
= 1012.58 − = 0.544 e +
!&
9×0.0604
∑) U ! 1.735 ≤ ≤ 23.59
" '(" 1' *1 &.NAA e+
m$*! = = = 0.0604 ⇒ 0.023 ≤ e + ≤ 0.311
$*! O
m$*! = 0.0604 = 0.246
Example
A random sample of size 20 from a normal distribution has a mean Y̅ =
32.8 and a standard deviation ]?/! = 4.51.
(a) Construct a 95% C.I. for S. (b) Construct a 95% C.I. for T.
Solution We have:
$ = 20, ]?/! = 4.51, &^ = 32.8, f = 0.05,
(a) kH/# $ − 1 = k*.*#- 19 = 2.093.
I
The C.I. is &^ ± kH/# $ − 1 0-,
?
,.-!
= 32.8 ± 2.093× = (30.69, 34.91)
#*
# #
(b) mH/# $ − 1 = m*.*#- 19 = 32.85
# #
m!/H/# $ − 1 = m*..J- 19 = 8.907
*
?/! K0-, *
?/! K0-, !.×,.-!* !.×,.-!*
#
The C.I. for T is * , * = ,
L=/* ?/! L,-=/* ?/! %#.'- '..*J
= 11.76, 43.39
The C.I. for T is 11.76, 43.39 = (3.430, 6.587)
 Summary
Assume that
1. !~` N, O " (Population is normally distributed)
2. !! , … , !$ is a random sample of size ^ from population !.
3. !X and m$*!" are the sample mean and sample variance.

/m − J
~ o 0,1
Q/ L

1 − f % C.I. on S (T # known)
T T
^
& − nH/# B , &^ + nH/# B
$ $
 Summary
Assume that
1. '~v T, l # (Population is normally distributed)
2. '$ , … , '; is a random sample of size % from population '.
3. '] and ^;2$#
are the sample mean and sample variance.
'] − T #
% − 1 ^;2$
~ b(% − 1) # ~ k# % − 1
^;2$ / % l
N O* = O+ + ⋯ + O, N O*& = O+& + ⋯ + O,&

&
1 1 1
OR = N O* &
T,-+ = N O*& − N O*
S S−1 S

1 − S % C.I. on T (l # unknown)
^;2$ ^;2$
'] − b?/# % − 1 x , '] + b?/# % − 1 x
% %
1 − S % C.I. on l #
# #
% − 1 ^;2$ % − 1 ^;2$
# , #
k?/# %−1 k$2?/# %−1