SPSS Statistics

SPSS Statistics

Your hub for statistical analysis, data management, and data documentation. Connect, learn, and share with your peers! 

 View Only

  • 1.  Binary logistic regression

    Posted Thu August 10, 2023 02:41 PM
      |   view attached

    Hello.

    I just bought a license and tried to use binary logistic regression but there is no button under the regression.

    if someone tells me how I can get the function, it would be greatly appreciated. 

    Sincerely,

    Sato



    ------------------------------
    Ryosuke Sato
    ------------------------------


  • 2.  RE: Binary logistic regression

    Posted Tue August 15, 2023 11:52 AM

    Hi @Ryosuke Sato

    Logistic Regression is a procedure found in the IBM SPSS Statistics Regression module.  This module is not present in the IBM SPSS Statistics Base Edition.  While it can be added to the Base edition, it is included by default in the IBM SPSS Statistics Standard, Professional, and Premium editions.  Please see the pricing page to review options.



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

    Original Message


  • 3.  RE: Binary logistic regression

    Posted Mon October 23, 2023 11:01 AM

    I had the same problem with the subscription. I did find that I could add on Firth Binary Logistic Regression. It's different but it does seem to produce usable results. I'm not sure when this subscription change came about, but now I am planning to research what the differences between the version we used to use and this one.



    ------------------------------
    Bettina Hansel
    ------------------------------

    Original Message


  • 4.  RE: Binary logistic regression

    Posted Mon October 23, 2023 11:26 AM
    If you run Firth logistic but choose regular (nonpenalized) maximum likelihood, you get the same answer as the built-in regular logistic regression.  But the underlying R package used for Firth requires that the dependent variable be set to a scale measurement level whether regular or penalized likelihood.  Otherwise it would come across to R as a factor rather than a numerical value.

    It would be natural to think of the DV as nominal or categorical, but that just not how the R package works, and the answer with the DV as scale does give the right answer.

    As far as I can recall, the behavior of Firth logistic has not changed.

    --



    Original Message


  • 5.  RE: Binary logistic regression

    Posted Mon October 23, 2023 11:46 AM

    Thank you, Jon Peck. That's very good to know. Previously I had used the binary logistic regression that no longer seems to be available on the menu in my subscription. Can you tell me how to get this back? When I looked for add-ons, I only found the Firth Logistic Regression. Looking into it, there seemed to be some advantages with smaller samples like I am working with now. I'll try it with a scale variable instead of a dummy variable, and see what the results are. The results I had using the dummy variable as a scale variable seemed credible, but I don't understand the Firth version the way I do the regular version. 



    ------------------------------
    Bettina Hansel
    ------------------------------

    Original Message


  • 6.  RE: Binary logistic regression

    Posted Mon October 23, 2023 11:55 AM
    Firth logistic is generally used for cases where regular logistic fails because of complete separation (perfect prediction).  It adds a penalty term to the likelihood equation.  Complete separation is more likely in small samples.

    As for the regular LOGISTIC REGRESSION command (Analyze > Regression > Binary Logistic), it is included in the Standard Edition as well as in the Regression option.  Run SHOW LIC or File > Manage License to see what you have. 

    --



    Original Message