SPSS Statistics

 View Only

  • 1.  Weighted least squares in hierarchical regression

    Posted Mon October 17, 2022 10:37 AM
    Reply

    Hi everyone, I've been looking for an answer to my question in this forum (and elsewhere in the internet) but haven't been able to find an answer. So I'm hoping someone can help me out on this:

    Is it possible to do a Weighted Least Square in Hierarchical Regression? With multiple IV in step/model 1 and/or 2? 


    If so, how to do this? Help is highly appreciated! Thank you, Paul



    ------------------------------
    Paul Schreuder
    ------------------------------

    #SPSSStatistics


  • 2.  RE: Weighted least squares in hierarchical regression

    Posted Mon October 17, 2022 10:47 AM
    Reply
    Hi.

    If you are referring to models such as those discussed in threads like this, then no, SPSS does not directly do that.

    That thread suggests software that does (I know nothing about it).

    I will send a note to the statisticians on our team and see what light they can shed.

    ------------------------------
    Rick Marcantonio
    Quality Assurance
    IBM
    ------------------------------

    Original Message


  • 3.  RE: Weighted least squares in hierarchical regression

    Posted Mon October 17, 2022 10:59 AM
    Reply
    Hi Rick, thanks for your quick reply. More specifically, I am wondering what weight to apply, because I would think that each model in an hierarchical regression has its own (best) weight estimate, but SPSS allows only for 1 (and the same) weight for both models in the procedure. Not sure if this is still an SPSS rather than statistical question, but finding an answer to it would really help me. Thank you! Paul

    ------------------------------
    Paul Schreuder
    ------------------------------

    Original Message


  • 4.  RE: Weighted least squares in hierarchical regression

    Posted Mon October 17, 2022 11:23 AM
    Reply
    It depends. If these are sampling weights, then the Complex Samples procedures (CSGLM) would be appropriate.  If this is a mixed model, then MIXED or GENLINMIXED might apply.  If these are heteroscedastic weights, then WLS or ordinary regression would work.

    --



    Original Message


Related Content