Hypothesis Tests

STAT*2090

Dr. Clémonell Bilayi-Biakana

Department of Mathematics and Statistics

University of Guelph

January 30, 2023.

Outline

Fundamental Concepts - Estimation

Fundamental Concepts - Hypothesis Tests

One-Sample Inference for a Population Mean

Two-Sample Inference for Population Means

“The only way to learn mathematics is to do

mathematics.” - Paul Halmos.
Hypothesis Testing: Two-Sample Inference
Normal Populations with Unknown Equal Variances

Let (X1 , X2 , . . . , Xn ) be a sample from X ⇠ N (µ1 , 12 ).

Let (Y1 , Y2 , . . . , Ym ) be a sample from Y ⇠ N (µ2 , 22 ).
Further, we assume that X ? ? Y and 12 = 22 = 2 , with 2 -
unknown. The sample sizes are assumed to be small. Consider
performing one of the following hypothesis tests:
1. H0 : µ1 = µ2 versus H1 : µ1 6= µ2 .
2. H0 : µ1 = µ2 versus H1 : µ1 > µ2 .
3. H0 : µ1 = µ2 versus H1 : µ1 < µ2 .
with a significance level ↵. Under H0 , the test statistic

X̄1 X̄2
T0 = q ⇠ Tn+m 2,
Sp2 (1/n + 1/m)
 in factD-V/rs,
t ButM( -

Nz
- -
x y
i -

N(V2, x
-y -
N(M 1,
-

2 e)F
+

sigmafactored x-y-N(M. -

Vz,5(t m) +

out

r2 <sp (n
=
1)si
- + (m -

Ds
17m - 2

x -

y -

(Ny(z) To
=

Sp t h +
Hypothesis Testing: Two-Sample Inference

where the pooled sample variance estimate is:

(n 1)s12 + (m 1)s22
sp2 =
n+m 2
Let t0 denote the observed value of T0 . is to
I-
=

1. If H0 : µ1 = µ2 versus H1 : µ1 6= µ2 , then reject H0 if +

|t0 | > tn+m 2;↵/2 with P T0 > tn+m 2;↵/2 = ↵/2 or
p-value= 2P (T0 > |t0 |) < ↵.

3
2. If H0 : µ1 = µ2 versus H1 : µ1 > µ2 , then reject H0 if
t0 > tn+m 2;↵ , with P (T0 > tn+m 2;↵ ) = ↵ or
f
p-value= P (T0 > t0 ) < ↵.
3. If H0 : µ1 = µ2 versus H1 : µ1 < µ2 , then reject H0 if
t0 < tn+m 2;↵ , with
P (T0 < tn+m 2;↵ ) = P (T0 > tn+m 2;↵ ) = ↵ or Enjo
p-value= P (T0 < t0 ) < ↵.
Remark
By inverting the acceptance region of the hypothesis testing
problem (1), we get a level 100(1 ↵)% C.I for µ1 µ2 :
r !
1 1
Iµ1 µ2 = (x̄1 x̄2 ) ± tn+m 2;↵/2 sp + .
n m
Exercise
Based on an indication that mean daily car rental rates may be
higher for Toronto than for Montreal, a survey of eight car rental
companies in Toronto is taken and the sample mean car rental rate
is 47, withastandarddeviationof 3. Further, suppose a survey of nine
car rental companies in Montreal results in a sample mean of
44andastandarddeviationof 3. Use = 0.05 to determine whether
the average daily car rental rates in Toronto are significantly higher
than those in Montreal. Assume car rental rates are normally
distributed and the population variances are equal.
 X:
Daily Car rental in Toronto

Y:Dailycar rental in Montreal

By assumptions,
C ANSr',ise 52-unknown

Further
S n 85 =47 Si 3
=
=

9% 445 3
=

m= =

we need to test (H):

Stories
Under Ho, To x y
=
-

T8 a Tis
=

2
+ -

spt t +

to x = -

y
sp]
+
47
=
-
44

(8 1(3)+(9 1)(3)2 X
b
-
-
+

89 + -
2

2.06
=

P-value approach
P-value (T. to) P(50 = 2.06) To -T15
=

>

it follows from the table that

1.753<2.06 <2.131
P(T. - 2.131) <P(7.>2.06) <P(T0> 1.793)
0.025 <P-value <0.05 =8-less then 2

P-value is less than Hence the Ho

Thus 1. we
reject
CriticalValue Approach
tn
we
rejectto to
if >
m
+ -

2, c,
with P(Totn m
+
-

2,c) 2
=

we have P(To +8+9-2,0.05) 0.05TorTis =

P(50T,5,0.05) 0.05 0.05 =

=

+15,0.05 1.753
=

7o
1.7533
Since 2.06
=

=> 70> +5
+ =

(5,0.05

Then the to thatis N.N2

we
reject
Hypothesis Testing: Two-Sample Inference
Normal Populations with Unequal Variances

Let (X1 , X2 , . . . , Xn ) be a sample from X ⇠ N (µ1 , 12 ).

Let (Y1 , Y2 , . . . , Ym ) be a sample from Y ⇠ N (µ2 , 22 ).
Further, assume that X ? ? Y and 12 6= 22 , both unknown.
Consider performing one of the following tests:
1. H0 : µ1 = µ2 versus H1 : µ1 6= µ2 .
2. H0 : µ1 = µ2 versus H1 : µ1 > µ2 .
3. H0 : µ1 = µ2 versus H1 : µ1 < µ2 .
with a significance level ↵. Under H0 , the test statistic

X̄ Ȳ
T0 = q 2 ⇡ T⌫ ,
S1 S22
n + m
Hypothesis Testing: Two-Sample Inference
Normal Populations with Unequal Variances

where ⌫ = min(n 1, m 1).The observed value of the test

statistic T0 is
x̄ ȳ
t0 = q 2 ⇡ T⌫ ,
s1 s22
n + m

1. If H0 : µ1 = µ2 versus H1 : µ1 6= µ2 , then reject H0

if |t0 | > t⌫;↵/2 , with P T0 > t⌫;↵/2 = ↵/2
or p-value= 2P (T0 > |t0 |) < ↵.
2. If H0 : µ1 = µ2 versus H1 : µ1 > µ2 , then reject H0 if
t0 > t⌫;↵ , with P (T0 > t⌫;↵ ) = ↵
or p-value= P (T0 > t0 ) < ↵.
3. If H0 : µ1 = µ2 versus H1 : µ1 < µ2 , then reject H0 if
t0 < t⌫;↵ , with P (T0 < t⌫;↵ ) = P (T0 > t⌫;↵ ) = ↵ or
p-value= P (T0 < t0 ) < ↵.
 X-N(n,E.)
in fact
Si -

(0 =)
Butv. -M2 > x-J:PointEstimate

xy N(n, y,z E)
-

=> -
-
+

After 5, 5
Estimating
T x y
=
-
-
(n, )-

Tu
=

Si 1 +
Remark
By inverting the acceptance region of the hypothesis testing
problem (1), we get a level 100(1 ↵)% C.I for µ1 µ2 :
r !
s12 s22
Iµ1 µ2 = (x̄1 x̄2 ) ± t⌫;↵/2 + .
n m
Exercise
There is a claim that the average blood volume in millimeters for
males who are paraplegic and participate in vigorous physical
activities is higher that that of males who are able-bodied and
participate in normal physical activities. A sample of size n1 = 7
paraplegic men gives x̄1 = 1511.714 and s12 = 49669.905. A
sample of size n2 = 10 able-bodied men gives x̄2 = 1118.400 and
s22 = 15297.600. Justify this claim using a 95% confidence interval
for µ1 µ2 . Justify the claim using the method of hypothesis
testing. Assume that the populations are normal with unequal
variances.
 solution:

of
X:Blood volume
Paraplegic men

1: Blood volume ofable bodied men

By assumptionx-N(M,5,2) > 5,152

y N(nz,52)
-

Further n = 7 ·X = 12 =

m 10
=

y = 5=

so 95% CI
a in M, -

r2 is

IN -

R (1 y
= - +
=
vz 5
+

Where P(T> tre) E, = TuTr

v
=

min (7 -1, 10-D min(6,0)

=

6
=

Since 1 -
2 =
0.93 =c = 0.05 =3 E =
0.025
Then P(T<+,0.025) 0.025
=

((1511.714
Thus In, -v 2.44)
1118.4 =
4663 +120
= -

I (66.053,620.575)
=

We are 95% confidentthatM.-N,30

N(> Nz
 Ho.vEe
2.
(H):

-
under H., To x y
=

S s
+

T To

v
=

min (7 -1, 10-D min(6,0)

=

6
=

-
To x y
=

S s
+

T To

393.314
=

44664.905
15247.67
+

7 10

4.234
=

Critical value Approach

we
reject to if tosto, a, with P(Toh,,0.0g) 0.05
=

to, 0.05 1.943-value from table

=

ection
Since4.23> 1.94
ronxes region

en
I

↑Manece.
then we do
reject to
 P-value Approach sided problem
Right
P-value=P(T.sto),To T,
=

P(T54.234)

The table yeilds

3.70774.234/4.317
P(T0> 4.3K) <P(T0> 4.234) < P(50 < 3.707)
0.0023 <P-value < 0.005 <a

RejectHo