SPSS Statistics

 View Only
Expand all | Collapse all

Please how do I interpret this Post Hoc Test After Kruskal Wallis H test

  • 1.  Please how do I interpret this Post Hoc Test After Kruskal Wallis H test

    Posted Mon March 20, 2023 09:26 AM

    Hi,

    I am trying to carry out a post hoc test after performing a Kruskal-Wallis test. I have 4 independent samples of 26 factors that am trying to compare. I used the Analyze> Non-Parametric Test> Independent samples..... path. The result showed that there are significant differences in about 6 factors for the four groups of the samples. The post hoc test was to know where exactly the significant differences occur among the 4 groups. I have attached images of the model viewer interface. The first image shows (on the left in yellow) where the significant differences occur, but the second image did not indicate where the significant differences occur despite that the Kruskal-Wallis test shows that there is a significant difference among the 4 groups regarding the factor. Please how do I interpret this output? The Kruskal-Wallis test shows that there is a significant difference among the 4 groups regarding the factor, but the post hoc is not showing exactly where the significant difference occur. Thank you. I appreciate your feedback.



    ------------------------------
    Anthony Yusuf
    ------------------------------



  • 2.  RE: Please how do I interpret this Post Hoc Test After Kruskal Wallis H test

    Posted Mon March 20, 2023 03:56 PM

    From one of our statisticians:

    "The overall or omnibus test is significant at the specified .05 level. The pairwise comparisons are shown in the table with individual significance or p values both unadjusted for multiple comparisons (second to last column on the right), and adjusted for multiple comparisons (last column). The highlighting of significant comparisons is done based on the multiplicity-adjusted p values, and none of those in your second table are below .05, so that's why none are highlighted. The third and fifth pairwise comparisons are significant at the .05 level if no multiplicity adjustment is made, but it is not actually always the case that a significant omnibus test will be associated with any significant pairwise comparisons, even if no adjustments for multiple comparisons are made."



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



  • 3.  RE: Please how do I interpret this Post Hoc Test After Kruskal Wallis H test

    Posted Tue March 21, 2023 07:30 PM
    Thank you. This is helpful. I appreciate it.
    Anthony Yusuf