SPSS Statistics

 View Only
  • 1.  Odd formatting in Regression Outlier Statistics Table

    Posted Sun June 26, 2022 03:07 PM
    Hi all,

    please consider the Statistic column in the attached table output, especially the Residual and Deleted Residual Rows.
    Any hints on what might cause this unexpected formatting?

    This is my syntax, applied to the Emplyee Data sample data file:

    REGRESSION
    /VARIABLES (COLLECT)
    /DESCRIPTIVES
    MEAN
    STDDEV
    CORR
    COV
    VARIANCE
    SIG
    N
    ALL
    /MISSING
    LISTWISE
    /STATISTICS
    DEFAULTS
    R
    COEFF
    ANOVA
    CHANGE
    COLLIN
    TOL
    SELECTION
    /CRITERIA TOLERANCE (.000001)
    /DEPENDENT salary
    /METHOD ENTER
    salbegin
    jobtime
    /RESIDUALS
    DEFAULTS
    DURBIN
    OUTLIERS (ZRESID RESID SRESID SDRESID DRESID MAHAL COOK)
    ID (id)
    NORMPROB (ZRESID PRED RESID ZPRED DRESID SRESID SDRESID)
    HISTOGRAM (ZRESID PRED RESID ZPRED DRESID ADJPRED SRESID SDRESID SEPRED MAHAL COOK LEVER)
    SIZE POOLED
    /CASEWISE
    OUTLIERS (1)
    PLOT (ZRESID)
    DEPENDENT
    PRED
    RESID
    ZRESID
    DRESID
    SRESID
    SDRESID
    /PARTIALPLOT ALL .

    ------------------------------
    Frank Watzl
    ------------------------------

    #SPSSStatistics


  • 2.  RE: Odd formatting in Regression Outlier Statistics Table

    Posted Fri July 01, 2022 01:42 PM

    Hi Frank,
    Using your command syntax and the 'Employee data.sav' sample file I was abler to replicate what you are seeing.  I did have some trouble with that command syntax however.

    - Copying and pasting from here to the Syntax Editor Window resulted in potential formatting issues (mainly correct indentation on continuation lines)
    - In the Syntax Editor, I saw several instances where values to Keywords were supplied but they did not appear colored correctly -- appearing as black and making me think these values are not acceptable for the given subcommand and keyword.
    - How did you create this command syntax?  I tried to recreate your syntax using the 'File -> Regression -> Linear' menus.  My result was markedly different from yours -- I had significantly fewer specifications..

    To be sure, the REGRESSION command probably has more specifications than can be created in the UI.  However, that's where I'd suggest you begin:
    - Setup as much of your analysis from the menus as you can and "Paste" what the UI has built.
    - Test that what has been pasted works
    - Begin adding specifications and testing them
    - The REGRESSION command is fully documented here: https://www.ibm.com/docs/en/spss-statistics/28.0.0?topic=reference-regression

    I hope this helps!



    ------------------------------
    David Dwyer
    ------------------------------



  • 3.  RE: Odd formatting in Regression Outlier Statistics Table

    Posted Fri July 01, 2022 02:27 PM
    Hi David,

    thank you for looking into this. The formatting of the table remains unchanged when I run the minimal example below. The syntax in my first example was "handwritten", as is the current minimal example. The first syntax was created during an exercise to get better acquainted with the REGRESSION command, and I readily admit that it was overloaded, possibly syntactically erroneous, and certainly not a realistic use case.

    The outlier table is created by the /RESIDUALS subcommand. I have not yet figured out how to generate this subcommand in the graphical user interface. As far as I can see, however, my usage should be compliant with the syntax documentation.

    REGRESSION
    /DEPENDENT = salary
    /METHOD = ENTER salbegin jobtime
    /RESIDUALS = OUTLIERS (RESID DRESID).





    ------------------------------
    Frank Watzl
    ------------------------------



  • 4.  RE: Odd formatting in Regression Outlier Statistics Table

    Posted Fri July 01, 2022 02:36 PM
    Edited by System Fri January 20, 2023 04:14 PM

    Hi Frank,
    I went back top the drawing board on this one and I see the problem is on the /VARIABLES subcommand.   The presence of the inappropriate 'gender' (a string) and 'bdate' (a date) variables is throwing off this table.  You should be able to explicitly mention your variables of interest on the /VARIABLES subcommand and have the REGRESSION procedure ignore those inappropriate fields -- but that doesn't work either.

    I've logged a defect with IBM SPSS Statistics  Development. 
    Please see Known Issue [DT139242](https://www.ibm.com/mysupport/aCI3p000000Xhzl)

    Thank you for bringing this to our attention!



    ------------------------------
    David Dwyer
    ------------------------------



  • 5.  RE: Odd formatting in Regression Outlier Statistics Table

    Posted Fri July 01, 2022 03:34 PM
    Hi David,

    thank you for following up.