Appropriate Bonferroni correction

#1
I'm comparing two groups of patients using several 3x2 and 4x2 contingency tables (for different sublevels of chemotherapy for instance). Question is how much do I need to adjust P value using Bonferroni correction (for 4 or 6 comparisons when analyzing 2x4 table, for 5 or 10 comparisons when analyzing 5x2 table). In this example I intend to find out do two groups differ in specific chemotherapies, Chemo 1 and Chemo 4 significantly differ, but what about Chemo 2?

Table looks like this:

Overall P <0.001

Variable / Group1 / Group2 / P value
Chemo 1 / 60 / 0 / <0.001
Chemo 2 / 40 / 1 / 0.011
Chemo 3 / 6 / 2 / 0.172
Chemo 4 / 1 / 8 / <0.001
 

hlsmith

Omega Contributor
#2
the conservative approach is correcting for each possible comparison. For example:


1 vs 2
1 vs 3
1 vs 4
2 vs 3
2 vs 4
3 vs 4


So multiply pvalues by 6.