SPSS Statistics

I have compared three groups along their means using the general linear model in SPSS. I also requested post hoc pairwise comparisons, and the p-values are sometimes equal to 1.000, which seems somewhat unlikely. I am wondering if this indicates that there is any error in the analyses or it is a consequence of 'Adjustment for multiple comparisons: Bonferroni', which note is associated with the column containing the p-values (on the other hand, in my understanding the Bonferroni correction affects the interpretation of p-values but not the values themselves).

Thanks for any input, much appreciated!

Barna

By definition, the control does not require any assumptions about dependence among the p-values or about how many of the null hypotheses are true.

Bonferroni also has the undesirable trait of false negatives.

There are many choices for post hoc tests. Bonferroni will certainly raise the p values, but it wouldn't account for a value of 1. That could happen with small n's or small differences.

If you double click the post hoc table and double click in a significance cell, you can see the sig level to more decimal places, but it could lrgitimately be 1.000

I have Excel so I suggest using that and see if the results are the same, could be that SPSS has problems

Thanks so much, Jon. I did try without the Bonferroni corrections and then I got more reasonable p-values.​