This page is hosted on AFS file server space, which is being shut down on November 13, 2018. If you are seeing this message, your service provider needs to take steps now. Visit afs.unc.edu for more information.

THE RUNS TESTS - AN ALTERNATIVE (OR ADDITIONAL) NONPARAMETRIC TEST OF AUTOCORRELATION OF RESIDUALS WITH THE DIVORCE DATA SET


>USE "D:\mydocs\ys209\divorce.syd"

SYSTAT Rectangular file D:\mydocs\ys209\divorce.syd,
created Wed Apr 19, 2000 at 00:25:57, contains variables:

 YEAR         DIVPOP       DIVMF        UNEMP        FLFPRT       MARUMF
 TREND

First do the regression and save the residuals (repeat).

>mglh
>select year <= 1970
>model divmf=constant+unemp+flfprt+marumf
>save divres01/data
>estimate

Data for the following results were selected according to:
      year <= 1970

Dep Var: DIVMF   N: 51   Multiple R: 0.866237   Squared multiple R: 0.750366
Adjusted squared multiple R: 0.734432   Standard error of estimate: 1.206846

Effect         Coefficient    Std Error     Std Coef Tolerance     t   P(2 Tail)
CONSTANT         -9.243243     2.117510     0.000000   .       -4.36515  0.00007
UNEMP             0.047519     0.035985     0.125499  0.588047  1.32052  0.19306
FLFPRT            0.239771     0.029453     0.672784  0.777664  8.14082  0.00000
MARUMF            0.137210     0.019332     0.606853  0.726515  7.09745  0.00000

                             Analysis of Variance

Source             Sum-of-Squares   df  Mean-Square     F-ratio       P
Regression            205.764749     3    68.588250   47.091854    0.000000
Residual               68.454466    47     1.456478

*** WARNING ***
Case           26 is an outlier        (Studentized Residual =     3.239025)
Case           27 has large leverage   (Leverage =     0.347000)
Case           27 is an outlier        (Studentized Residual =     3.955823)

Durbin-Watson D Statistic     0.429
First Order Autocorrelation   0.711

Residuals have been saved.
-------------------------------------------------------------------------------

Plot of Residuals against Predicted Values


Second, use the file of residuals, go to the npar (nonparametric statistics) module, and do the runs test.  A significant result indicates that the residuals are correlated

>use divres01

SYSTAT Rectangular file d:\mydocs\ys209\divres01.SYD,
created Wed Apr 26, 2000 at 13:29:59, contains variables:

 ESTIMATE     RESIDUAL     LEVERAGE     COOK         STUDENT      SEPRED
 YEAR         DIVPOP       DIVMF        UNEMP        FLFPRT       MARUMF
 TREND

>npar
>runs residual
 

Wald-Wolfowitz runs test using cutpoint =        0.000
                                                                   Probability
 Variable    Cases LE Cut  Cases GT Cut   Runs         Z              (2-tail)
 RESIDUAL          32            19         8     -5.103269          0.000000

Thus one concludes that the residuals in this case are correlated.



Last modified 26 Apr 2000