Sei sulla pagina 1di 8

STA2020F

Class Test 2
Date: 29 April 2009
Time: 1 hr and 15 mins
Instructions: Answer all questions in your answer book. The appropriate tables and a
formula sheet are supplied.
Question 1.
The following data come from the standings of teams competing in a sports league.
e wish to determine if there is a negati!e relationship between the number of points
scored by each team "#$% and the number of points scored against "#A% during an
eight game period.
Team &'$ (A) *+, $A$ -&( $.- A)&
#A "/% 119 205 101 100 101 123 134
#$ "y% 14 9 3 3 4 0 2
,i!en5 6/ 7 1 114 6/
2
7 131 512 6y 7 51 6y
2
7 521 6/y 7 2 925
a% (alculate the appropriate test statistic for testing the null hypothesis that / and y are
uncorrelated against the alternati!e that they are negati!ely correlated.
"1%
b% 8ow perform the test9 gi!ing a p:!alue "as far as is possible% in your conclusion.
"1%
c% .ind the least squares equation for predicting points scored by a team gi!en the
points scored against.
"2%
[8]
Question 2.
A professor wants to test if the number of hours of study and the number of e/ercises
completed during the course by a student has an effect on the mark of the student in
the course. ;e considers 12 students and fits a first order multiple regression model
with two independent !ariables "numbers of hours of study and e/ercises completed%.
The following is obtained5
Source DF SS MS F
Reression 20
!rror
Total 30
a% .ind and interpret the coefficient of determination.
"2%
b% $tate the null and alternati!e hypotheses9 along with a p:!alue and your conclusion9
for the <o!erall= or <global= test of the model.
"4%
[8]
Question ".
>ata were collected on the weekly gas consumption "1000 cubic feet% for a home in
+ngland. The a!erage outside temperature "degrees .ahrenheit% for each week is also
recorded. There are 00 measurements in total9 with 24 weeks of data before insulation
was installed and 13 weeks of data after insulating. The house thermostat was set at
the same le!el throughout.
a% The following output was obtained from fitting a simple linear regression model
with y 7 gas consumption and / 7 a!erage outside temperature5
#redictor (oef $+ t #:!al
(onstant 5.1291 0.2022 21.94 0.000
Temp :0.21402 0.00221 :0.51 0.000
$ 7 0.3511 *:$q 7 12.3? *:$q"ad@% 7 11.2?
Analysis of Aariance
$ource >. $$ B$ . #:!al
*egression 1 10.912 10.912 20.03 0.000
+rror 02 10.523 0.223
Total 01 05.090
'nterpret the coefficients of the fitted model. (omment on the fit of the model.
"0%
b% (learly gas consumption will be affected by the absence or presence of insulation
in the house. +/plain how the !ariable <'nsul= can be included in the model.
"1%
c% The following output was obtained from fitting a multiple regression model with
gas consumption as the dependent and <Temp= and <'nsul= as the independent
!ariables5
#redictor (oef $+ t #:!al
(onstant 4.2121 0.1149 52.03 0.000
Temp :0.14249 0.01339 :19.02 0.000
'nsul :1.2904 0.1015 :12.11 0.000
$ 7 0.2991 *:$q 7 91.9? *:$q"ad@% 7 91.5?
Analysis of Aariance
$ource >. $$ B$ . #:!al
*egression 2 01.313 20.909 211.03 0.000
+rror 01 1.422 0.090
Total 01 05.090
'nterpret the coefficients of the fitted model. (omment on the fit of the model.
"5%
d% hat is meant by interaction in the conte/t of regressionC ;ow in practice can an
interaction term be included in a statistical modelC
"1%
e% The following output was obtained from fitting a multiple regression model with
gas consumption as the dependent and <Temp=9 <'nsul= and <TempD'nsul= "the
interaction% as the independent !ariables5
#redictor (oef $+ t #:!al
(onstant 4.3513 0.1114 40.12 0.000
Temp :0.19120 0.01329 :20.91 0.000
'nsul :2.2412 0.1223 :11.10 0.000
TempD'nsl 0.10141 0.00055 1.22 0.001
$ 7 0.2499 *:$q 7 91.4? *:$q"ad@% 7 91.1?
Analysis of Aariance
$ource >. $$ B$ . #:!al
*egression 1 02.525 10.192 190.22 0.000
+rror 00 2.915 0.021
Total 01 05.090
"i% #redict the gas consumption of the house when insulation is present and the
a!erage outdoor temperature is 0.0.
"ii% rite a paragraph reporting your findings and any conclusions that may be drawn
from fitting this final model.
"3%
[2"]
Question #.
The graph below was constructed as part of a multiple regression problem5
a% hat does this graph tell you about the !alidity or otherwise of the assumptions on
which regression is basedC EThe !ertical a/is shows absolute !alue of residuals9 not
residuals themsel!es.F
"0%
b% 'f you decide that not all the assumptions abo!e are met9 what can be done about itC
"2%
[$]
STA2020F 200%: Im&ortant Formulae
A'()A:
Source of Variation SS df MS F F
crit
&etween groups $$T k G 1 B$T B$THB$+ .
I9 k:1. n:k
ithin groups $$+ n G k B$+
T(TA* SS+Total, n - 1
Source of
Variation
SS df MS F F
crit
*ows $$& b G 1 B$& B$&HB$+ .
I9 b:1. n:k:bJ1
(olumns $$T k G 1 B$T B$THB$+ .
I9 k:1. n:k:bJ1
+rror $$+ n G k G b J 1 B$+
T(TA* SS+Total, n - 1
Source of Variation SS df MS F F
crit
$ample $$"&% b G 1 B$"&% B$"&%HB$+ .
I9 b:1. n:ab
(olumns $$"A% a G 1 B$"A% B$"A%HB$+ .
I9 a:1. n:ab
'nteraction $$"A&% "a G 1%"b G 1% B$"A&% B$"A&%HB$+ .
I9"a:1%" b:1%. n:ab
ithin $$+ n G ab B$+
T(TA* SS+Total, n - 1
Fis.er/s *SD:
H 29
1 1
n k
i j
t MSE
n n

_
+


,
0on1erroni A23ustment:
H 2 9
1 1
9 " 1% H 2
E
C n k
i j
t MSE C k k
n n

_
+


,
Tu4e5/s Critical rane5
9 9
1 1
2
k
i j
MSE
q
n n

_
+


,
where n k
SIM6*! *I'!AR R!7R!SSI(':
Measures o1 association:
( ) ( ) ( )
( )
2
2
2

co!" 9 %
1
xy i i x i
xy xy
x y
x
x y
SS x x y y xy SS x x x
n n
SS SS
X Y r
n
SS SS



(*S estimation:
Testin t.e correlation coe11icient: Stan2ar2 error o1 estimate:
6re2iction inter8al:
2
29 2
2
" %
1
K 1
" 1%
g
n
x
x x
y t s
n n s

t + +

Con1i2ence inter8al:
2
29 2
2
" %
1
K
" 1%
g
n
x
x x
y t s
n n s

t +

M9*TI6*! R!7R!SSI(' :
Stan2ar2 error o1 estimate:
1
SSE
s
n k



Coe11icient o1 multi&le 2etermination:
[ ]
2
2 2
2 2 2
co!" 9 %
1
" %
x y i
X Y
SSE
R or R
s s y y

A23uste2 R
2
:
2
H" 1%
1
" % H" 1%
i
SSE n k
y y n

1
1 0 1
2
2
1
" %" %
co!" 9 %

" %
n
i i
xy
i
n
x x
i
i
x x y y
SS
X Y
b b y b x
SS s
x x

2
2
df 7 2
1 2
n SSE
t r n s
r n



Coe11icient o1 &artial 2etermination:
R F
R
SSE SSE
SSE


6artial F test:
" % H
R F
F
SSE SSE r
MSE

Reression A'()A ta:le:


df SS MS F Sig F
*egression k $$* B$* 7 $$*Hk B$*HB$+ #".
k9 n:k:1
L .%
*esidual n:k:1 $$+ B$+ 7 $$+H"n:k:1%
T(TA* n-1 SST
T-test statistic:
df 7 1
i
i i
b
b
t n k
s


'('-6ARAM!TRIC:
Runs Test +normal a&&ro;< n1 = 20 > n2 = 20,:
R
R
R
z

?.ere
1 2
1 2
2
1
R
n n
n n
+
+
<
2 1 2 1 2 1 2
2
1 2 1 2
2 "2 %
" % " 1%
R
n n n n n n
n n n n

+ +
0inomial Test: e/act p:!alue @
n
i n i
i k
n
p q
i

_

,

0inomial Test +normal a&&ro;< n = 2A,:

?.ere

np 9

npq
C.i sBuare 7oo2ness-o1-Fit Test:
2
2
e/p
i
i
obs
! n

1 df k
Cilco;on Ran4 Sum Statistic +normal a&&ro;< n = 10,:

?.ere
1 1 2 1 2 1 2
" 1% " 1%
9
2 12

n n n n n n n

+ + + +

Cilco;on Sine2 Ran4 Sum Statistic +normal a&&ro;< n = "0,:

?.ere
" 1% " 1%"2 1%

0 20

n n n n n

+ + +

Sin Test Statistic +normal a&&ro;< n = 2A,:
H 2
H 0
S n
z
n

Mc'emar/s Test o1 S5mmetr5:


2
2
" % " C
!
" C

+
1 df
Drus4al Callis Statistic:
2
1
12
1" 1%
" 1%
k
j
j
j

# n
n n n

1
+
1
+
1
]

1 df k
Frie2man Statistic:
2
1
12
1 " 1%
" 1%
k
r j
j
F b k
bk k

1
+
1
+
]


1 df k
S&earman Ran4 correlation +n E "0,:
2
1
2
4
1
" 1%
n
i
i
s
d
r
n n


S&earman Ran4 correlation +normal a&&ro;< n = "0,: 1
s
$ r n
C.i sBuare Test o1 Association:
2
2
e/p
i
i
obs
! n

" 1%" 1% df r c
7amma Test:
2

"1 %
% !
& &
z '
& '
+

?.ere


% !
% !
& &
'
& &

+
TIM! S!RI!S:
!;&onential smoot.in: $
t
7 "w%y
t
J "1 : w%$
t:1
for t L 1 9 0
M w M 1
Tren2 !Buation: .
t
7 b
0
J b
1
t
Tren2 !Buation ?it. Seasonal In2ices5 .
t
7 "b
0
J b
1
t% N $'
t
Dur:in Catson:
2
1
2
2
1
" %
n
t t
t
n
t
t
e e
d
e

2
n
t
1 t 1
O O "P %
BA> B$+
n
t t t
t
Y F F
n n

Potrebbero piacerti anche