SPSS Statistics

 View Only
Expand all | Collapse all

Where is the R-version recognised by SPSS 27 on a MAC (OS Big Sur - V11.2.3)

  • 1.  Where is the R-version recognised by SPSS 27 on a MAC (OS Big Sur - V11.2.3)

    Posted Mon March 15, 2021 03:35 PM
    Help needed
    I run SPSS V27 for MAC (macOS Big Sur), annual subscription, and I am trying to associate SPSS with R.
    Despite many attempts I did not find a way to find upload the R version that works with SPSS V27 on MAC (OS BigSur - V11.2.3).
    R 3.6 is recommended  but when R 3.6 is installed on MAC, SPSS does not recognize this version.
    Could you provide me with the pathway to download the right R version working with SPSS V27 for MAC?
    I need this extension for running propensity analysis on SPSS.
    Thanks,
    JBR


    ------------------------------
    Jean-Baptiste Ricco
    ------------------------------

    #SPSSStatistics


  • 2.  RE: Where is the R-version recognised by SPSS 27 on a MAC (OS Big Sur - V11.2.3)

    IBM Champion
    Posted Mon March 15, 2021 03:44 PM
    You need R3.6.x.  Did you install the R36_CONFIGURATION file and then run it from the Extensions menu?

    Once that is done, you can install the PSMatching file if that is the one  you want to run, but I suggest that you get the version from here
    as it fixes a problem with the dialog box that causes trouble otherwise.

    There is also a Data > Propensity Score Matching custom dialog available within Statistics without the need for R if that suits your needs.

    --





  • 3.  RE: Where is the R-version recognised by SPSS 27 on a MAC (OS Big Sur - V11.2.3)

    Posted Thu March 25, 2021 01:37 PM
    I am not using the new Apple chip but the old one with a 2,3 GHz 8-Core Intel Core i9 and MacOs 11.2.3
    Concerning the propensity score analysis (PSA) by SPSS without R implementation, the least I can say is that it is incomplete without any possibility to calliper each covariate included in the score separately but as a whole, which is not appropriate when you have continuous variables together with categorical variables.
    Example: One variable is "age" in years and you choose a calliper of 1 (one year) assuming a difference of one year, which is fine, if another variable is "diabetes" with 0 as "no diabetes" and 1 as "diabetes", keeping calliper=1 is a nonsense. Following this, you have no choice but choosing a "global" calliper of 0, same "age" exactly which limits the number of patients following PSA to a few.
    I did this on a sample of 530 patients with two treatment groups and 6 covariates, SPSS (without R), and a calliper =0, finds only 24 patients available following PSA, On the same sample, same variables, PSA with Rstudio and matchIt or other libraries (not SPSS) find between 120 and 144 patients available following matching with a well balanced repartition of variables of interest in each group. As PSA is applied more and more to compare treatments when randomisation is not possible, SPSS should be interested into this.

    ------------------------------
    Jean-Baptiste Ricco
    ------------------------------