Regression with Autocorrelated Errors
U.S. Wine Consumption and Adult Population – 1934-2002
Data Description
• Y=U.S. Annual Wine Consumption (Millions of Gallons)
• X=U.S. Adult Population (Millions of People)• Years – 1934-2002 (Post Prohibition)• Model:
2
11
10
,0~
...
iid
XY
t
qtqttt
ttt
Ordinary Least Squares Regression
64.4828.236631.497.347^^
ett PW
Regression StatisticsMultiple R 0.965612383R Square 0.932407274Adjusted R Square 0.931398427Standard Error 48.64438813Observations 69
ANOVAdf SS MS F Significance F
Regression 1 2186985.91 2186985.91 924.23 0.0000Residual 67 158540.53 2366.28Total 68 2345526.43
Coefficients Standard Error t Stat P-value Lower 95% Upper 95%Intercept -347.9736 21.9895 -15.8245 0.0000 -391.8649 -304.0824apop_m 4.3092 0.1417 30.4012 0.0000 4.0263 4.5921
present isation Autocorrel
64.170,158.170,1
0.11989871158540.53
19008.8044 :TestWatson -Durbin
1
2
2
21
U
UL
n
tt
n
ttt
dDW
npdnpd
e
eeDW
Covariance Structure (q=1)
122
21
2
1
2
22
1
12
2
2
2
2
2
112121
2
2
2
2222
2
222
12
222
2
2
11
2110
,1
1
1
1
1
,
1,
10
:GeneralIn
1,,
11
1
10
10 :Assuming
11,0~
tttt
nn
n
n
n
t
ttk
ktttt
ttttttt
COVVV
VV
V
COVCOVVE
VCOVCOV
VVVE
VE
iidXY
ε
Generalized Least Squares (q=1)
εTXβTYTY*IVTTYTI
VTT
VT
TεYV
1111'11
1'1
1
1
Transform
1000
0100
0010
0001
1
1000
100
010
001
1000
1100
1110
1111
1
1000
1100
1110
1111
1
1
1
1
1
1
100
010
001
0001
:4 whereCase heConsider t
1000
01000
0010
0001
00001
:Define
1
1
1
1
22
2
2
2
2
2
2
2
2
22
2222
232222
2
2
2
22
2222
232222
2
2
23
2
2
32
2
2
2
2
21
2
1
2
2
V
n
VV
nn
n
n
Estimated GLS (q=1)
i
iii
iEGLS
iEGLSie
e
n
ttt
n
tt
SEtsSE
pns
SEtV
pn
een
en
^
^^
2^
^^^
2
,
^
,
^1^1^2^^
1^1^
2^1^1^1^1^
^
^
^
^
^
2^
1^
^^^2^
^
^^
21
^
1
2^
)0(/1'
)1()0(
'
1'
''''
10000
01000
00010
00001
00000)0(
)1()0()0(
)1(1)1(
1)0(
1
EGLS
^
EGLS
^
EGLS
^
1
EGLS
^
XTTX'β
βXYTTβXYYTTX'XTTX'β
T
Estimated GLS (q=1) – Wine/Population Data
38.2104372.0
9348.004372.)0(/39.4
358.90.4546
4.2540
2066.0
4.2540
0.206631.515-
31.515-5482.225
5479.8024.2540
347.23-
10.9348-0000
010.9348-000
00010.9348-0
000010.9348-
000003552.0
82.8920.93482147.89)1(2297.69)0(
^
^^
2^
2
1
^
2^
1^
2^^^^
i
iii
e
SEtsSEs
tV
EGLS
^
EGLS
^
β
β
T
present ision tocorrelatconcludeAunot Do
64.170,158.170,167.1
U
UL
dDW
npdnpdDW
Estimated GLS (General q) - I
)()(
1000000
0100000
0000010
0000001
)()(
:))1(2 (assuming :matrixation transformObtain the 5)
: ofion DecompositCholesky Obtain the 4)
')0(
: and vector 1 - errors lagged theof tsCoefficien theEstimate 3)
)(
)2(
)1(
)0()2()1(
)2()0()1(
)1()1()0(
,...,1,0)(
: vector1 andmatrix following theand lag toancesautocovari theEstimate 2)
:Residuals OLSObtain and Regression SquaresLeast Ordinary Fit 1)
1
^^1
^^
1
^^1
^^
2^
11
^2^1
2
^
^
^
^^^
^^^
^^^
1^
nnnqn
nqqnqqq
qn
Vqx
qqq
q
q
qhn
eeh
qxqxqq
q
q
q
q
t
n
hthtt
2
1^
2
1211112q
^
11
1/2^^
q
^
q
^
q
^
q
^
q
^^
q
^
q
^^
q
^
q
^
OLS
^
OLS1
OLS
^
T
TTT
TTT0TPT
VT
P'PΓΓ
γργΓρ
γΓ
βXYeYX'XX'β
Estimated GLS (General q) - II
qpndfSE
t
isSEqpn
s
SEt
V
qpn
i
ii
thiiiii
iEGLS
iEGLSi
e
e
'on with distributi- t with thecompare and
: for the statistics-Obtain t 11)
ofelement diagonal '
')0(
:errors standard theand parameters siveautoregres of estimatesfor MS residualObtain 10)
:tscoefficien regressionfor statistics- testimatedObtain 9)
': ofmatrix covariance- varianceestimated Obtain the 8)
'
''
:model ed transform thefrom anceerror vari Estimated Obtain the 7)
'':estimate squaresleast dgeneralize estimated Obtain the 6)
^
^
i
1^^2
^
^
2
,
^
,
^
2^^
2^
q
^q
^^
1^^
EGLS
^
EGLS
^
EGLS
^^^
EGLS
^
^^1^^
EGLS
^
Γγρ
XTTX'ββ
βXYTTβXY
YTTX'XTTX'β
SAS Proc Autoreg Output The AUTOREG Procedure Dependent Variable wine Ordinary Least Squares Estimates SSE 158540.525 DFE 67 MSE 2366 Root MSE 48.64439 SBC 738.318203 AIC 733.84999 Regress R-Square 0.9324 Total R-Square 0.9324 Durbin-Watson 0.1199
Standard Approx Variable DF Estimate Error t Value Pr > |t| Intercept 1 -347.9736 21.9895 -15.82 <.0001 adpop 1 4.3092 0.1417 30.40 <.0001
Estimates of Autocorrelations Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1
0 2297.7 1.000000 | |********************| 1 2147.9 0.934807 | |******************* |
SAS Proc Autoreg Output
Preliminary MSE 289.8
Estimates of Autoregressive Parameters Standard Lag Coefficient Error t Value 1 -0.934807 0.043717 -21.38
Yule-Walker Estimates SSE 18516.1612 DFE 66 MSE 280.54790 Root MSE 16.74956 SBC 596.454422 AIC 589.752103 Regress R-Square 0.5702 Total R-Square 0.9921 Durbin-Watson 1.6728
Standard Approx Variable DF Estimate Error t Value Pr > |t|
Intercept 1 -347.2297 74.0420 -4.69 <.0001 adpop 1 4.2540 0.4546 9.36 <.0001