SPSS Statistics

Greetings! I have a syntax file that I run to generate a Spearman partial correlation.

NONPAR CORR covar1 covar2 var1 var2 var3
    /MISSING = LISTWISE
/MATRIX OUT(*) .

RECODE rowtype_('RHO'='CORR') .

PARTIAL CORR var1 var2 var3 BY covar1 covar2
    /SIGNIFICANCE = TWOTAIL
    /MISSING = LISTWISE
    /MATRIX IN(*) .

(full credit to Steven Bradburn: How To Perform A Non-Parametric Partial Correlation In SPSS)

After pasting the output into Excel, it looks something like this (minus var3 for simplicity, values made up):

This normally works very well and only requires minor cleanup for self-comparisons. However, I've run into an issue where adding another covariable (covar4) with no other changes results in an output like this:

I am stuck trying to figure out why this is happening. If it helps, all variables involved are numeric. Covar1 and 4 are scales, while covar2 and 3 are nominal.

Any help figuring this out would be greatly appreciated.

I do not see much sense so far in doing this. Why not simply to rank data first and then run Partial Correlations procedure on those data? Spearman is just Pearson correlation computed on ranks, and Partial Correlations expects specifically Pearson r coefficients. There is no thing like "nonparametric partial correlation" in all this - but parametric partial correlations based on data that are the ranks.

------------------------------
Kirill Orlov
------------------------------

NONPAR CORR covar1 covar2 var1 var2 var3    /MISSING = LISTWISE/MATRIX OUT(*) .RECODE rowtype_('RHO'='CORR') .PARTIAL CORR var1 var2 var3 BY covar1 covar2    /SIGNIFICANCE = TWOTAIL    /MISSING = LISTWISE    /MATRIX IN(*) .