1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
|
@c PSPP - a program for statistical analysis.
@c Copyright (C) 2017, 2020 Free Software Foundation, Inc.
@c Permission is granted to copy, distribute and/or modify this document
@c under the terms of the GNU Free Documentation License, Version 1.3
@c or any later version published by the Free Software Foundation;
@c with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.
@c A copy of the license is included in the section entitled "GNU
@c Free Documentation License".
@c
@node REGRESSION
@section REGRESSION
@cindex regression
@cindex linear regression
The @cmd{REGRESSION} procedure fits linear models to data via least-squares
estimation. The procedure is appropriate for data which satisfy those
assumptions typical in linear regression:
@itemize @bullet
@item The data set contains @math{n} observations of a dependent variable, say
@math{Y_1,@dots{},Y_n}, and @math{n} observations of one or more explanatory
variables.
Let @math{X_{11}, X_{12}}, @dots{}, @math{X_{1n}} denote the @math{n} observations
of the first explanatory variable;
@math{X_{21}},@dots{},@math{X_{2n}} denote the @math{n} observations of the second
explanatory variable;
@math{X_{k1}},@dots{},@math{X_{kn}} denote the @math{n} observations of
the @math{k}th explanatory variable.
@item The dependent variable @math{Y} has the following relationship to the
explanatory variables:
@math{Y_i = b_0 + b_1 X_{1i} + ... + b_k X_{ki} + Z_i}
where @math{b_0, b_1, @dots{}, b_k} are unknown
coefficients, and @math{Z_1,@dots{},Z_n} are independent, normally
distributed @dfn{noise} terms with mean zero and common variance.
The noise, or @dfn{error} terms are unobserved.
This relationship is called the @dfn{linear model}.
@end itemize
The @cmd{REGRESSION} procedure estimates the coefficients
@math{b_0,@dots{},b_k} and produces output relevant to inferences for the
linear model.
@menu
* Syntax:: Syntax definition.
* Examples:: Using the REGRESSION procedure.
@end menu
@node Syntax
@subsection Syntax
@vindex REGRESSION
@display
REGRESSION
/VARIABLES=@var{var_list}
/DEPENDENT=@var{var_list}
/STATISTICS=@{ALL, DEFAULTS, R, COEFF, ANOVA, BCOV, CI[@var{conf}, TOL]@}
@{ /ORIGIN | /NOORIGIN @}
/SAVE=@{PRED, RESID@}
@end display
The @cmd{REGRESSION} procedure reads the active dataset and outputs
statistics relevant to the linear model specified by the user.
The @subcmd{VARIABLES} subcommand, which is required, specifies the list of
variables to be analyzed. Keyword @subcmd{VARIABLES} is required. The
@subcmd{DEPENDENT} subcommand specifies the dependent variable of the linear
model. The @subcmd{DEPENDENT} subcommand is required. All variables listed in
the @subcmd{VARIABLES} subcommand, but not listed in the @subcmd{DEPENDENT} subcommand,
are treated as explanatory variables in the linear model.
All other subcommands are optional:
The @subcmd{STATISTICS} subcommand specifies which statistics are to be displayed.
The following keywords are accepted:
@table @subcmd
@item ALL
All of the statistics below.
@item R
The ratio of the sums of squares due to the model to the total sums of
squares for the dependent variable.
@item COEFF
A table containing the estimated model coefficients and their standard errors.
@item CI (@var{conf})
This item is only relevant if COEFF has also been selected. It specifies that the
confidence interval for the coefficients should be printed. The optional value @var{conf},
which must be in parentheses, is the desired confidence level expressed as a percentage.
@item ANOVA
Analysis of variance table for the model.
@item BCOV
The covariance matrix for the estimated model coefficients.
@item TOL
The variance inflation factor and its reciprocal. This has no effect unless COEFF is also given.
@item DEFAULT
The same as if R, COEFF, and ANOVA had been selected.
This is what you get if the /STATISTICS command is not specified,
or if it is specified without any parameters.
@end table
The @subcmd{ORIGIN} and @subcmd{NOORIGIN} subcommands are mutually
exclusive. @subcmd{ORIGIN} indicates that the regression should be
performed through the origin. You should use this option if, and
only if you have reason to believe that the regression does indeed
pass through the origin --- that is to say, the value @math{b_0} above,
is zero. The default is @subcmd{NOORIGIN}.
The @subcmd{SAVE} subcommand causes @pspp{} to save the residuals or predicted
values from the fitted
model to the active dataset. @pspp{} will store the residuals in a variable
called @samp{RES1} if no such variable exists, @samp{RES2} if @samp{RES1}
already exists,
@samp{RES3} if @samp{RES1} and @samp{RES2} already exist, etc. It will
choose the name of
the variable for the predicted values similarly, but with @samp{PRED} as a
prefix.
When @subcmd{SAVE} is used, @pspp{} ignores @cmd{TEMPORARY}, treating
temporary transformations as permanent.
@node Examples
@subsection Examples
The following @pspp{} syntax will generate the default output and save the
predicted values and residuals to the active dataset.
@example
title 'Demonstrate REGRESSION procedure'.
data list / v0 1-2 (A) v1 v2 3-22 (10).
begin data.
b 7.735648 -23.97588
b 6.142625 -19.63854
a 7.651430 -25.26557
c 6.125125 -16.57090
a 8.245789 -25.80001
c 6.031540 -17.56743
a 9.832291 -28.35977
c 5.343832 -16.79548
a 8.838262 -29.25689
b 6.200189 -18.58219
end data.
list.
regression /variables=v0 v1 v2 /statistics defaults /dependent=v2
/save pred resid /method=enter.
@end example
|
