HYPOTHESIS TESTING 2
X
where t ~ t (n 1)
s
n
and s is the sample standard deviation. t has a student tdistribution and has degree of freedom (d.f.=n-1).
If is not known, use t test as t =
i =1
standard deviation S = S 2
2. t distribution: What does the t distribution look like?
a. Similar to the z distribution: Symmetric around zero, Mound
shaped, and Mean is zero.
b. The t distribution has wider tails
c. There is a different distribution for each degree of freedom
value.
d. What determines how much wider is the t-distribution than
the normal distribution?
t .025,df 10 = 2.228
t .025,df 28 = 2.048
t .025,df 100 = 1984
.
t .025,df = 196
.
X
If the sample size is small, n<30, s has a t distribution if
n
X
s
n
is
has a t distribution
is approximately normally
II. t test
1. t test: confidence interval
Confidence interval for t-test
=
( x t 2
s
s
, x + t 2
)
n
n where d.f.=n-1
( x z
, x + z
Example:
Unoccupied seats on flights cause airlines to lose revenue.
Suppose a large airline wants to estimate its average number of
unoccupied seats per flight over the past year. To accomplish
this, the records of 16 flights are randomly selected, ant the
number of unoccupied seats is noted for each of the sampled
flights. The sample mean is 11.6. Estimate , the mean number
of unoccupied seats per flight during the past year using a 95%
confidence interval.
a. The population standard deviation is known as 5 seats.
b. The population standard deviation is unknown and the
sample standard deviation is known as 5 seats.
a. z test
=0.05. /2=0.025.
z 2 = z0.025 = 1.96
Confidence interval:
( x z 2
, x + z 2
) = (11.6 1.96
5
5
,11.6 + 1.96
) = (11.6 2.45,11.6 + 2.45)
16
16
b. t test
=0.05. /2=0.025.
2, df = n 1
Confidence interval:
( x t
s
, x + t
n
s
5
5
) = (11.6 2.131
,11.6 + 2.131
) = (11.6 2.66,11.6 + 2.66)
n
16
16
X 28,000 25,000
=
= 2.32
s
5,000
t .01, 14
n
= 2.624
15
E( X1 X 2 ) = E( X1) E( X 2 ) = 1 2
V ( X1 X 2 ) = V ( X1) + V ( X 2 ) =
12
n
22
n
( X1 X 2 ) Z
2
12
n1
( X 1 X 2 ) ( 1 2 )
12
n1
22
n2
22
n2
2.
2.
( X1 X 2 ) t
2
1
1
+ )
n1 n2
S p2 (
1
1
+ )
n1 n2
d. f.= n 1 + n2 2
S p2
( n1 1) s12 + ( n2 1) s22
=
n 1 + n2 2
s12 s22
+
n1 n2
( X1 X 2 ) t
2
(X 1 X 2 ) ( 1 2 )
d. f .=
s12 s22
+
n1 n2
s1
n1
22
s1
n
1
n1 1
s
+ 2
n2
n2 1
s2
2
n
2
PART III.
Testing the Population Proportion
We use sample proportions to estimate the population
proportion
p =
x
n
p(1 p)
provided
n
z=
p p
p(1 p)
n
Example:
In the fall of 1998, a newspaper publisher launched a new
national newspaper in Canada. It was believed that the new
newspaper would have to capture at least 12% of the Toronto
market in order to be financially viable. During the planning
stages of this news newspaper, a market survey was conducted
of a sample for 400 Toronto readers. After providing a brief
description of the proposed newspaper, one question asked if
the survey participants would subscribe to the newspaper if the
cost did not exceed $20 per month. If 58 participants said that
they would subscribe, can the publisher conclude that the
proposed newspaper will be financially viable? Let =.05
H 0 : p =.12
H1: p >.12
p p
.145.12
z=
=
= 154
.
p(1 p)
.12(1.12)
n
400
p value = P( Z > 154
. ) =.0618
Rejection Region starts at 1.645. Cannot reject the null
hypothesis.
EXCEL EXAMPLES
HYPOTHESIS TESTING
Formulate hypotheses
For each of the following statements, formulate appropriate
null and alternative hypotheses and indicate whether the
appropriate test will be one-tail or two-tail.
a. The average TOEFL score for international students is
well above 575.
b. The average American adult drinks less than 3 cups of
coffee per day
c.The average housewife works more than 40 hours per
week in house related activities.
d. The average employee calls in sick 3 times a year.
z-test
The Admissions officer for the graduate programs at
Michigan State University (MSU) believes that the average
score on the GRE exam at his university is significantly
higher than the national average of 1300. Assume that the
population standard deviation is 125 and a random sample
of 25 scores had an average of 1375.
a. State the appropriate null and alternative hypotheses
b. Calculate the value of the test statistic and set up the
rejection region. What is your conclusion?
c.Calculate the p-value
d. Does the p-value confirm the conclusion in part b?
Computer examples
Standard normal distribution
NORMSDIST(z) = Pr(Z < z)
NORMSDIST(1.96) = 0.975
NORMSINV(0.975) = 1.96
t-distribution
TDIST(t,df,tails) gives us p-value
TDIST(1.96,1000,1)=0.025
TDIST(1.96,1000,2)=0.05
TINV(alpha, df)
TINV(0.05, 1000) = 1.96
=0.05
H0: = 170.
H1: > 170 (one-tail test)
2.
Recall p-value
>
178
178 170
) = P ( Z > 2.46) = 0.0069.
65
400