Information Technology Reference
In-Depth Information
Parm.
Est.
Parm.
Est.
Parm.
Est.
Parm.
Est.
-7.80
-.41
(
JT
)
-.15
(82b)
q
.53
(
CC
)
18
6( 4)
39
7
60
4
-5.69
(
CC
)
0
(
JT
)
-.28
(82b)
q
.14
19
7( 4)
40
8
61
5
-2.34
-.11
-.62
.23
(
CC
)
(
JT
)
(82b)
q
20
1(5)
41
9
62
6
-.045
(
CC
)
.092
(
JT
)
0
(82b)
q
0
21
2(5)
42
10
63
7
.11
-1.63
1.92
1.27
(
RC
)
(
CC
)
(79a)
q
(82d)
q
1
1
22
3(5)
43
1
64
1
(
RC
)
.59
(
CC
)
.11
(79a)
q
2.09
(82d)
q
.92
2
2
23
4(5)
44
2
65
2
(
RC
)
1.23
0
(79a)
q
1.43
(82d)
q
.67
(
CC
)
3
3
24
5(5)
45
3
66
3
(
RC
)
-.13
(
CC
)
-.13
(79a)
q
.84
(82d)
q
.77
4
4
25
1(6)
46
4
67
4
.59
.34
.58
.44
(
RC
)
(
CC
)
(79a)
q
(82d)
q
5
5
26
2(6)
47
5
68
5
(
RC
)
.40
(
CC
)
-.23
(79a)
q
.19
(82d)
q
.19
6
6
27
3(6)
48
6
69
6
0
0
0
0
(
RC
)
(
CC
)
(79a)
q
(82d)
q
7
7
28
4(6)
49
7
70
7
(
CC
)
-.45
(
CC
)
.42
(82a)
q
2.68
(82f)
q
.85
8
1(1)
29
1(7 )
50
1
71
1
-.60
0
(82a)
q
.71
(82f)
q
1.69
(
CC
)
(
CC
)
9
2(1)
30
2(7 )
51
2
72
2
(
CC
)
0
(
BC
)
-.21
(82a)
q
1.05
(82f)
q
1.08
10
3(1)
31
1
52
3
73
3
0
0
.89
.69
(
CC
)
(
BC
)
(82a)
q
(82f)
q
11
1(2 )
32
2
53
4
74
4
(
CC
)
0
(
JT
)
-.097
(82a)
q
.31
(82f)
q
.59
12
1(3)
33
1
54
5
75
5
.78
-.25
.14
.16
(
CC
)
(
JT
)
(82a)
q
(82f)
q
13
1( 4)
34
2
55
6
76
6
(
CC
)
-.53
(
JT
)
-.20
(82a)
q
0
(82f)
q
0
14
2( 4)
35
3
56
7
77
7
.83
(
JT
)
-.48
(82b)
q
1.31
(
CC
)
14
3( 4)
36
4
57
1
(
CC
)
-.050
(
JT
)
.11
(82b)
q
.81
16
4( 4)
37
5
58
2
.24
-.47
.75
(
CC
)
(
JT
)
(82b)
q
17
5( 4)
38
6
59
3
4
Table 2. Maximum Likelihood Estimates of
{ i
and 
1
The section of the PROC LOGISTIC output entitled 'Type-3 Analysis of Effects' characterizes
the statistical significance of the covariates through p-values obtained by referencing a Wald
chi-square test statistic to a corresponding null chi-square distribution. Table 3 shows the
chi-square tests and the corresponding p-values, and it is seen that all covariate groups are
highly significant contributors in the model.
One way to assess model adequacy for multinomial logistic regression is to use the model to
predict Y and then examine how well the predicted values match the true values of Y. Since
the output of the model for each customer is an estimated probability distribution for Y, a
natural predictor of Y is the mode of this distribution. We note that this predictor considers
equal cost for all forms of prediction errors. More elaborate predictors could be derived by
assuming a more complex cost model where, for example, the cost of predicting 5 when the
actual value is 1 is higher than the cost of predicting 5 when the actual value is 4. Table 4,
the so-called confusion matrix of the predictions, displays the cross classification of all 5056
customers based on their actual value of Y and the model-predicted value of Y.
Search WWH ::




Custom Search