SPSS Statistics

 View Only
  • 1.  SPSS Weighted Least Squares Regression Model

    Posted Mon March 01, 2021 06:18 PM

    When building a regression, I used SPSS Weighted Least Squares Regression to adjust for heteroscedaticity.

     

    When building the WEIGHT variable, does SPSS allow me to include multiple layers and multiple independent variables. See below.

     

     

    *COMMENT To generate the unstandardized residuals.

     

    REGRESSION

      /MISSING LISTWISE

      /STATISTICS COEFF OUTS R ANOVA

      /CRITERIA=PIN(.05) POUT(.10)

      /NOORIGIN

      /DEPENDENT Holistic

      /METHOD=ENTER Outside

      /METHOD=ENTER HoursWeekReading2 HoursWeekListening2 HoursWeekConversing2

      /SAVE RESID.

     

    *COMMENT Set Absolute value for AbsRES.

     

    COMPUTE AbsRES=abs(RES_1).

    EXECUTE.

     

    *Comment This generates the predicted residuals.

     

    REGRESSION

      /MISSING LISTWISE

      /STATISTICS COEFF OUTS R ANOVA

      /CRITERIA=PIN(.05) POUT(.10)

      /NOORIGIN

      /DEPENDENT AbsRES

      /METHOD=ENTER Outside

      /METHOD=ENTER HoursWeekReading2 HoursWeekListening2 HoursWeekConversing2

      /SAVE PRED.

     

    *COMMENT Create Weight variable.

     

    COMPUTE WEIGHT=1/(PRED_1**2).

    EXECUTE.

     

    *COMMENT Run the Weighted Least Squares Model.

     

    REGRESSION

      /MISSING LISTWISE

      /REGWGT=WEIGHT

      /STATISTICS COEFF OUTS R ANOVA

      /CRITERIA=PIN(.05) POUT(.10)

      /NOORIGIN

      /DEPENDENT Holistic

      /METHOD=ENTER Outside

      /METHOD=STEPWISE HoursWeekReading HoursWeekListening HoursWeekConversing.

     



    ------------------------------
    Stafford Cox
    ------------------------------

    #SPSSStatistics


  • 2.  RE: SPSS Weighted Least Squares Regression Model

    IBM Champion
    Posted Mon March 01, 2021 06:31 PM
    You construct the weight variable to correct for heteroscedasticity outside the regression procedure.  Regression does not impose  any constraints on this process.  It's up to you how to model the weights.

    You might find the SPSSINC BREUSCH PAGAN extension command helpful in modeling and testing for heteroscedasticity.  It requires R and the Rplugin and can be installed from the Extensions > Extension Hub menu if you don't already have it. No R skills are required to use it.

    --