SPSS Statistics

SPSS Statistics

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

 View Only
  • 1.  Predicted Probabilities and Log Likelihood on General Linear Mixed Models

    Posted Thu August 03, 2023 11:28 AM

    Hello everyone,

    I`m using a binary logistic in the General Linear Mixed Model sub-menu. I need the values for the Predicted Probabilities for each predictor (not case) and Log Likelihood of the regression model. Is it possible to obtain these values with SPSS? How can I get them or calculate them.

    Thanks for your help!



    ------------------------------
    Ricardo Carvalho
    ------------------------------


  • 2.  RE: Predicted Probabilities and Log Likelihood on General Linear Mixed Models

    Posted Thu August 10, 2023 07:42 AM

    Hi Richardo

     If you are looking into the mixed models, then click on Analyze > Mixed Models > Generalized Linear Mixed Models. After selecting options for Data Structure, Fields & Effects and Build Options, Click on "Model Options > Save Fields", you can see the option for "Predicted Probabilities for Categorical Targets".



    ------------------------------
    Bindu Krishnan
    ------------------------------



  • 3.  RE: Predicted Probabilities and Log Likelihood on General Linear Mixed Models

    Posted Sat August 12, 2023 03:16 AM

    Hi Bindu,

    Thanks for you reply and informatio! :)

    In my post I was referring to the calculation of predicted proabiblities per variable. That is: the mixed model logistic regression model makes predictions on a log odds scale, and I want convert this to a probability scale. Does the SPSS make this operation automatically?

    And does it calculate the Log-Likelihood fo the model? Or it calculates only the -2LL?

    Best regards,

    Ricardo



    ------------------------------
    Ricardo Carvalho
    ------------------------------



  • 4.  RE: Predicted Probabilities and Log Likelihood on General Linear Mixed Models

    Posted Mon August 14, 2023 04:57 AM

    Hi Ricardo

    In the output of binary logistic regression, the model summary includes -2log likelihood. In addition, in mixed models logistic regression, you will get its related measures of information criteria such as Akaike corrected, Bayesian , etc. Regarding the predicted probabilities, I would like to know the situation or the data that you are working on. Because usually in logistic regression, predicted probabilities are calculated for the prediction of category of target (or dependent) variable. So here, are you focussing on feature selection or some other ?



    ------------------------------
    Bindu Krishnan
    ------------------------------