SPSS Statistics

 View Only
  • 1.  Copy & Paste Bug in SPSS Workbooks

    Posted Tue March 26, 2024 10:18 AM

    I believe I may have discovered a bug in the SPSS Workbooks option.  In the classic environment, once a procedure is run, you can use the keyboard Ctrl+C to copy and then paste the results into Excel for further calculations.  When doing the same thing in the Workbook environment, the keyboard Ctrl+C action does not copy over to Excel.  Instead, you have to [Right-click] Copy As > Plain Text.  This is not practical when running many models, and need to use the workbook option to help manage heavy use of SPSS.  Is this a true bug, and if not, is there a way while working in the Workbook environment, to use the keyboard Ctrl+C action to more quickly copy plain text results into Excel for further processing?



    ------------------------------
    Mark Sloan
    ------------------------------


  • 2.  RE: Copy & Paste Bug in SPSS Workbooks

    Posted Tue March 26, 2024 05:03 PM

    Hi @Mark Sloan

    What version of IBM SPSS Statistics and which procedure are you copying output?
    For me, IBM SPSS Statistics 29.0.2.0 (20) copies and pastes the same DESCRIPTIVES pivot table output  the same way in Output Viewer and in Workbook; on Windows and on macOS.



    ------------------------------
    David Dwyer
    SPSS Technical Support
    IBM Software
    ------------------------------



  • 3.  RE: Copy & Paste Bug in SPSS Workbooks

    Posted Wed March 27, 2024 12:17 PM
    Edited by Mark Sloan Wed March 27, 2024 12:22 PM

    I am using SPSS 29.0.1 remotely from a university server (Georgia Tech), and the procedure is MIXED syntax, in the Workbook environment.  I am finding some similar cut and paste issues in SPSS 27.0.0 (64-bit) (classic only), where I must use the right-click menu copy because the keyboard Ctrl+C doesn't work.  This leads me to another issue, where MIXED converges in SPSS 27.0.0, but does not in SPSS 29.0.1. Most of the results are the same with convergence or not, but with a few differences, such as the denominator df showing different values (matching a textbook) in SPSS 27, but all show the same value in SPSS 29 (not matching the textbook). I'm a bit concerned with SPSS 29 at this point, as I would like to confidently use it moving forward.



    ------------------------------
    Mark Sloan
    ------------------------------



  • 4.  RE: Copy & Paste Bug in SPSS Workbooks

    Posted Thu March 28, 2024 09:25 AM

    Hi Mark,

    Thank you for providing such a detailed description of your experience with SPSS 29.0.1 and the differences you've encountered compared to SPSS 27.0.0. It sounds like you've been dealing with some challenges, particularly regarding convergence and discrepancies in results between the two versions.

    To assist you better, could you kindly share the syntax and/or data you've been working with in SPSS 29.0.1? Having access to this information will allow me to investigate the issue further and provide you with more targeted assistance. Additionally, if there are any specific analyses or tasks you're attempting to perform with SPSS 29.0.1, please feel free to share those as well.

    Thank you once again for reaching out, and I look forward to assisting you further.



    ------------------------------
    [Aruna Saraswathy,] [Statistician]
    [SPSS Statistics]
    [IBM]
    ------------------------------



  • 5.  RE: Copy & Paste Bug in SPSS Workbooks

    Posted Thu March 28, 2024 11:09 AM
      |   view attached

    Aruna,  Thank you for getting back to me.  I've attached the data file and below is the syntax that converges in SPSS 27, but does not converge in SPSS 29 (I do get a Hessian Matrix warning in both).  I really want to use SPSS 29 for its Workbook feature, but need to ensure convergence in version 29 is comparable to version 27.  Note the spaces in the syntax below are due to an automated input system I've developed in Excel -- SPSS doesn't seem to have a problem with it:

    TITLE '     TWO-LEVEL RANDOM COEFFICIENTS (SLOPE) MODEL for Garson lq2002.sav    '.                            
    MIXED    HOSTILE    WITH    TSIG    GTSIG    LEAD                
      /CRITERIA=DFMETHOD(        Satterthwaite    ) CIN(95)     MXITER(    100    ) MXSTEP(    10    ) SCORING(1)     
    SINGULAR(    0.000000000001    ) HCONVERGE(    0    , ABSOLUTE)                     
    LCONVERGE(    0    , ABSOLUTE) PCONVERGE(        0.000001    , ABSOLUTE)                
      /FIXED=    TSIG    GTSIG    LEAD         | SSTYPE(3)                
      /METHOD=    ML                                
      /PRINT=HISTORY(1) SOLUTION TESTCOV                                    
      /RANDOM=INTERCEPT         TSIG    | SUBJECT(    COMPID    ) COVTYPE(    VC    ).        



    ------------------------------
    Mark Sloan
    ------------------------------

    Attachment(s)

    sav
    lq2002.sav   169 KB 1 version


  • 6.  RE: Copy & Paste Bug in SPSS Workbooks

    Posted Mon April 01, 2024 07:09 AM

    Hi Mark,

    Thanks for sharing your syntax! I've reviewed it and have a couple of questions that might help us troubleshoot the convergence issue.

    1. Have you had a chance to try this syntax with other datasets? Understanding how it performs with different data can provide further insights.

    2. Collinearity between independent variables is a possibility we can explore. This occurs when independent variables are highly correlated. In mixed models with random effects, collinearity can lead to redundant covariance parameters, making it difficult for the model to estimate them accurately. Would you be able to check the correlations between your independent variables? High correlations (above 0.8 or 0.9) might indicate collinearity.

    To help investigate this further, I've included the same code modified for a different dataset below:


    TITLE '     TWO-LEVEL RANDOM COEFFICIENTS (SLOPE) MODEL for bankloan.sav    '.                            
    MIXED    default   WITH    debtinc   othdebt    income                
      /CRITERIA=DFMETHOD(        SATTERTHWAITE    ) CIN(95)     MXITER(    100   ) MXSTEP(    10    ) SCORING(1)     
    SINGULAR(    0.000000000001    ) HCONVERGE(    0    , ABSOLUTE)                     
    LCONVERGE(    0    , ABSOLUTE) PCONVERGE(        0.000001    , ABSOLUTE)                
      /FIXED=    debtinc   othdebt    income         | SSTYPE(3)                
      /METHOD=    ML                                
      /PRINT=HISTORY(1) SOLUTION TESTCOV                                    
      /RANDOM= INTERCEPT othdebt    | SUBJECT(    employ    )COVTYPE(VC) .    

    Please let me know if you have any questions about collinearity or if you'd like to explore other possibilities. I'm happy to help in any way I can.

    Best regards,



    ------------------------------
    [Aruna Saraswathy,] [Statistician]
    [SPSS Statistics]
    [IBM]
    ------------------------------



  • 7.  RE: Copy & Paste Bug in SPSS Workbooks

    Posted Mon April 01, 2024 03:07 PM
    Aruna, Thanks!  I am currently out of town so to answer #1, I have not tried this specifically with other data sets. Most I am testing are sample sets that have passed muster with the experts using them as examples.  For #2, I will need to test when I get the next chance. I do know that in SPSS 27 the three IVs did not prevent the model from converging so it appears multicollinearity was not an issue for SPSS 27. 

    Thank you for the follow-up.
    Mark






  • 8.  RE: Copy & Paste Bug in SPSS Workbooks

    Posted Fri April 05, 2024 09:45 AM

    I re-tested a model where SPSS 29 would not converge, but SPSS 27 did converge.  I closed out of SPSS 29 (workbook, data set, and program).  After a successful SPSS 27 convergence, I restarted SPSS 29 fresh with a new workbook, and the same model converged successfully!  I am wondering if there might be a bug in SPSS 29 MIXED with workbooks when SPSS is used heavily over a period of time.  Is there a standard way to reset/fresh-start SPSS 29 MIXED without closing out the workbook or data set?

    PS, In addition to my testing data set, I also successfully tested bankloan.sav with your syntax above in SPSS 29 with workbooks, and all worked as it should.  Multicollinearity was not an issue with the either data set.  I did find an issue with one of the dichotomous variables in my previous non-convergence (showed 0 in Number of parameters in MIXED Model Dimensions), that then showed up as a 1 in number of parameters after a fresh start with new workbook.  It appears to me there may be a bug in SPSS 29 workbooks after an extended period of use.  If there is a reset mechanism or syntax statement I can add, it would be greatly appreciated!

    Thank you for the follow-up.

    Mark



    ------------------------------
    Mark Sloan
    ------------------------------