-AT_BANNER([REGRESSION])
+AT_BANNER([LINEAR REGRESSION])
-AT_SETUP([REGRESSION])
+AT_SETUP([LINEAR REGRESSION - basic])
AT_DATA([regression.sps], [dnl
-data list list / v0 to v2.
+set format = F22.3.
+data list notable list / v0 to v2.
begin data
0.65377128 7.735648 -23.97588
-0.13087553 6.142625 -19.63854
list.
])
-AT_CHECK([pspp -o pspp.csv regression.sps])
-AT_CHECK([cat pspp.csv], [0], [dnl
-Table: Reading free-form data from INLINE.
-Variable,Format
-v0,F8.0
-v1,F8.0
-v2,F8.0
-
+AT_CHECK([pspp -O format=csv regression.sps], [0], [dnl
Table: Model Summary (v2)
,R,R Square,Adjusted R Square,Std. Error of the Estimate
-,.97,.94,.93,1.34
+,.971,.942,.925,1.337
Table: ANOVA (v2)
,,Sum of Squares,df,Mean Square,F,Significance
-,Regression,202.75,2,101.38,56.75,.00
-,Residual,12.50,7,1.79,,
-,Total,215.26,9,,,
+,Regression,202.753,2,101.376,56.754,.000
+,Residual,12.504,7,1.786,,
+,Total,215.256,9,,,
Table: Coefficients (v2)
,,B,Std. Error,Beta,t,Significance
-,(Constant),2.19,2.36,.00,.93,.38
-,v0,1.81,1.05,.17,1.72,.12
-,v1,-3.43,.33,-1.03,-10.33,.00
+,(Constant),2.191,2.357,.000,.930,.380
+,v0,1.813,1.053,.171,1.722,.129
+,v1,-3.427,.332,-1.026,-10.334,.000
,,,,,,
Table: Data List
v0,v1,v2,RES1,PRED1
-.65,7.74,-23.98,-.84,-23.13
--.13,6.14,-19.64,-.54,-19.10
-.35,7.65,-25.27,-1.87,-23.40
-.69,6.13,-16.57,.97,-17.54
--.07,8.25,-25.80,.40,-26.20
--.34,6.03,-17.57,1.53,-19.10
-.76,9.83,-28.36,1.77,-30.13
--.47,5.34,-16.80,.18,-16.97
--.06,8.84,-29.26,-1.05,-28.21
-.56,6.20,-18.58,-.54,-18.04
+.654,7.736,-23.976,-.84,-23.13
+-.131,6.143,-19.639,-.54,-19.10
+.349,7.651,-25.266,-1.87,-23.40
+.692,6.125,-16.571,.97,-17.54
+-.074,8.246,-25.800,.40,-26.20
+-.344,6.032,-17.567,1.53,-19.10
+.760,9.832,-28.360,1.77,-30.13
+-.470,5.344,-16.795,.18,-16.97
+-.061,8.838,-29.257,-1.05,-28.21
+.562,6.200,-18.582,-.54,-18.04
])
AT_CLEANUP
+
+# Test to ensure that the /SAVE subcommand works properly when SPLIT is active
+AT_SETUP([LINEAR REGRESSION - SAVE vs SPLITS])
+
+# Generate some test data based on a linear model
+AT_DATA([gen-data.sps], [dnl
+set seed = 1.
+input program.
+loop #c = 1 to 20.
+ compute x0 = rv.normal (0,1).
+ compute x1 = rv.normal (0,2).
+ compute err = rv.normal (0,0.1).
+ compute y = 4 - 2 * x0 + 3 * x1 + err.
+ compute g = (#c > 10).
+ end case.
+end loop.
+end file.
+end input program.
+
+print outfile='regdata.txt' /g x0 x1 y err *.
+execute.
+])
+
+AT_CHECK([pspp -O format=csv gen-data.sps], [0], [ignore])
+
+# Use our test data to create a predictor and a residual variable
+# for G == 0
+AT_DATA([regression0.sps], [dnl
+data list notable file='regdata.txt' list /g x0 x1 y err *.
+
+select if (g = 0).
+
+regression
+ /variables = x0 x1
+ /dependent = y
+ /statistics = all
+ /save = pred resid.
+ .
+
+print outfile='outdata-g0.txt' /g x0 x1 y err res1 pred1 *.
+execute.
+])
+
+
+AT_CHECK([pspp -O format=csv regression0.sps], [0], [ignore])
+
+# Use our test data to create a predictor and a residual variable
+# for G == 1
+AT_DATA([regression1.sps], [dnl
+data list notable file='regdata.txt' list /g x0 x1 y err *.
+
+select if (g = 1).
+
+regression
+ /variables = x0 x1
+ /dependent = y
+ /statistics = all
+ /save = pred resid.
+ .
+
+print outfile='outdata-g1.txt' /g x0 x1 y err res1 pred1 *.
+execute.
+])
+
+
+AT_CHECK([pspp -O format=csv regression1.sps], [0], [ignore])
+
+# Use our test data to create a predictor and a residual variable
+# The data is split on G
+AT_DATA([regression-split.sps], [dnl
+data list notable file='regdata.txt' list /g x0 x1 y err *.
+
+split file by g.
+
+regression
+ /variables = x0 x1
+ /dependent = y
+ /statistics = all
+ /save = pred resid.
+ .
+
+print outfile='outdata-split.txt' /g x0 x1 y err res1 pred1 *.
+execute.
+])
+
+AT_CHECK([pspp -O format=csv regression-split.sps], [0], [ignore])
+
+# The concatenation of G==0 and G==1 should be identical to the SPLIT data
+AT_CHECK([cat outdata-g0.txt outdata-g1.txt | diff outdata-split.txt - ], [0], [])
+
+AT_CLEANUP
+
+
+# Test that the procedure behaves sensibly when presented with
+# multiple dependent variables
+AT_SETUP([LINEAR REGRESSION multiple dependent variables])
+AT_DATA([regression.sps], [dnl
+set seed = 2.
+input program.
+loop #c = 1 to 200.
+ compute x0 = rv.normal (0, 1).
+ compute x1 = rv.normal (0, 2).
+ compute err = rv.normal (0, 0.8).
+ compute y = 2 - 1.5 * x0 + 8.4 * x1 + err.
+ compute ycopy = y.
+ end case.
+end loop.
+end file.
+end input program.
+
+regression
+ /variables = x0 x1
+ /dependent = y ycopy
+ /statistics = all
+])
+
+AT_CHECK([pspp -O format=csv regression.sps > output], [0], [ignore])
+
+
+AT_CHECK([head -16 output > first], [0], [])
+AT_CHECK([tail -16 output > second], [0], [])
+
+AT_CHECK([sed -e 's/ycopy/y/g' second | diff first -], [0], [])
+
+
+AT_CLEANUP
+
# Tests the QR decomposition used by the REGRESSION command.
-AT_SETUP([REGRESSION test of QR decomposition])
+AT_SETUP([LINEAR REGRESSION test of QR decomposition])
AT_DATA([regression.sps], [dnl
data list list / v0 to v1.
begin data
])
AT_CLEANUP
-AT_SETUP([REGRESSION no crash on all missing])
+AT_SETUP([LINEAR REGRESSION no crash on all missing])
AT_DATA([regcrash.sps], [dnl
data list list /x * y.
begin data.
-AT_SETUP([REGRESSION missing dependent variable])
+AT_SETUP([LINEAR REGRESSION missing dependent variable])
dnl Test for a bug where missing values in the dependent variable were not being
dnl ignored like they should have been.
AT_CLEANUP
-AT_SETUP([REGRESSION with invalid syntax (and empty dataset)])
+AT_SETUP([LINEAR REGRESSION with invalid syntax (and empty dataset)])
AT_DATA([ss.sps], [dnl
data list notable list / v0 to v2.
projects / pspp / blobdiff