Re: SPSS warning from regression
Posted by Richard Ristow on Mar 08, 2007; 9:55pm
URL: http://spssx-discussion.165.s1.nabble.com/SPSS-warning-from-regression-tp1074315p1074317.html
At 12:56 PM 3/8/2007, Don Asay wrote:
>Load all the variables you want to look at into a bivariate
>correlation. Under Analyze --> Correlate --> Bivariate. If you see a
>box full of periods, your
>regression won't work.
Actually, if you see ONE period your regression won't work. The period
(correlation missing) means that the two variables are valid in two or
fewer cases. The correlation is a normalized form of the covariance
needed for regression; so the covariance can't be computed either.
HOWEVER, you can see no periods at all, and it still won't work. With
listwise variable deletion (which you should use), only cases with all
variables present are available for the regression. You can see this
happen, if you run CORRELATIONS with listwise deletion. Like this (SPSS
15 draft output; WRR-not saved separately):
List
|-----------------------------|---------------------------|
|Output Created |08-MAR-2007 16:45:08 |
|-----------------------------|---------------------------|
CASENUM DEPEND IND1 IND2 IND3
1 .9 . 9.8 7.9
2 7.9 4.4 . 6.8
3 3.7 5.5 1.4 .
4 5.7 . 5.6 7.4
5 9.0 3.3 . 8.0
6 1.8 8.7 8.2 .
7 9.5 . 2.7 8.9
8 6.3 6.0 . 5.4
9 5.6 5.2 3.8 .
Number of cases read: 9 Number of cases listed: 9
CORRELATIONS
/VARIABLES=DEPEND IND1 IND2 IND3
/PRINT=TWOTAIL NOSIG
/MISSING=PAIRWISE .
Correlations
|-----------------------------|---------------------------|
|Output Created |08-MAR-2007 16:45:08 |
|-----------------------------|---------------------------|
Correlations
|------|---------------|--------|--------|-----|-----|
| | |DEPEND |IND1 |IND2 |IND3 |
|------|---------------|--------|--------|-----|-----|
|DEPEND|Pearson |1 |-.900(*)|-.689|.135 |
| |Correlation | | | | |
| |---------------|--------|--------|-----|-----|
| |Sig. (2-tailed)| |.015 |.130 |.799 |
| |N |9 |6 |6 |6 |
|------|---------------|--------|--------|-----|-----|
|IND1 |Pearson |-.900(*)|1 |.903 |-.993|
| |Correlation | | | | |
| |---------------|--------|--------|-----|-----|
| |Sig. (2-tailed)|.015 | |.283 |.073 |
| |N |6 |6 |3 |3 |
|------|---------------|--------|--------|-----|-----|
|IND2 |Pearson |-.689 |.903 |1 |-.563|
| |Correlation | | | | |
| |---------------|--------|--------|-----|-----|
| |Sig. (2-tailed)|.130 |.283 | |.619 |
| |N |6 |3 |6 |3 |
|------|---------------|--------|--------|-----|-----|
|IND3 |Pearson |.135 |-.993 |-.563|1 |
| |Correlation | | | | |
| |---------------|--------|--------|-----|-----|
| |Sig. (2-tailed)|.799 |.073 |.619 | |
| |---------------|--------|--------|-----|-----|
| |N |6 |3 |3 |6 |
|------|---------------|--------|--------|-----|-----|
* Correlation is significant at the 0.05 level (2-tailed).
CORRELATIONS
/VARIABLES=DEPEND IND1 IND2 IND3
/PRINT=TWOTAIL NOSIG
/MISSING=LISTWISE .
Correlations
|-----------------------------|---------------------------|
|Output Created |08-MAR-2007 16:45:08 |
|-----------------------------|---------------------------|
Correlations(b)
|------|---------------|------|----|----|----|
| | |DEPEND|IND1|IND2|IND3|
|------|---------------|------|----|----|----|
|DEPEND|Pearson |.(a) |.(a)|.(a)|.(a)|
| |Correlation | | | | |
| |---------------|------|----|----|----|
| |Sig. (2-tailed)| |. |. |. |
|------|---------------|------|----|----|----|
|IND1 |Pearson |.(a) |.(a)|.(a)|.(a)|
| |Correlation | | | | |
| |---------------|------|----|----|----|
| |Sig. (2-tailed)|. | |. |. |
|------|---------------|------|----|----|----|
|IND2 |Pearson |.(a) |.(a)|.(a)|.(a)|
| |Correlation | | | | |
| |---------------|------|----|----|----|
| |Sig. (2-tailed)|. |. | |. |
|------|---------------|------|----|----|----|
|IND3 |Pearson |.(a) |.(a)|.(a)|.(a)|
| |Correlation | | | | |
| |---------------|------|----|----|----|
| |Sig. (2-tailed)|. |. |. | |
|------|---------------|------|----|----|----|
a Cannot be computed because at least one of the variables is constant.
b Listwise N=0
REGRESSION
/MISSING LISTWISE
/STATISTICS COEFF OUTS R ANOVA
/CRITERIA=PIN(.05) POUT(.10)
/NOORIGIN
/DEPENDENT DEPEND
/METHOD=ENTER IND1 IND2 IND3 .
Regression
|-----------------------------|---------------------------|
|Output Created |08-MAR-2007 16:46:39 |
|-----------------------------|---------------------------|
Warnings
|--------------------------------------------------------|
|There are no valid cases for models with dependent |
|variable DEPEND. |
|Statistics cannot be computed. |
|--------------------------------------------------------|
|No valid cases found. |
|Equation-building skipped. |
|--------------------------------------------------------|
===================
APPENDIX: Test data
===================
NEW FILE.
INPUT PROGRAM.
. NUMERIC CASENUM (F3).
. NUMERIC DEPEND (F5.1).
. NUMERIC IND1 TO IND3 (F5.1).
. LOOP CASENUM = 1 TO 9.
. VECTOR DATA_VAR=DEPEND TO IND3.
. LOOP #VNUM = 1 TO 4.
. COMPUTE DATA_VAR(#VNUM) = RV.UNIFORM(0,10).
. COMPUTE #ZAP = 2 + MOD((CASENUM-1),3).
. COMPUTE DATA_VAR(#ZAP) = $SYSMIS.
. END LOOP.
END CASE.
. END LOOP.
END FILE.
END INPUT PROGRAM.
LIST.