Post hoc test following Fisher's exact test

Hi. I performed Fisher’s exact test in SPSS on a sample of 141 patients with diabetic foot. The rows are comorbidity grades (zero, one, two, three), while the columns are diabetic foot prognosis (No amputation, Amputation). and as shown in the image I got a significant difference (P value 0.010). My main question is what post hoc test to use after Fisher’s exact test to determine which cell or pair of cells are responsible for the significance in the data? My other questions are what does the value 10.401 shown immediately beside the sentence “Fisher’s exact test” represent? and what does the value 3.041 in the sentence “The standardized statistic is 3.041” represent? I would be grateful for any help

A few general remarks:

  • You are not using the ordinal nature of comorbidity. It’s very easy to do so were it to have been the dependent variable, not so easy for an indendent variable. To approximate a better way I often treat such variables as quadratic in a regression model.
  • By not embedding the problem into a logistic regression model you’re missing out on some opportunities such as adjusting for other variables and getting simultaneous confidence intervals.
  • Fisher’s “exact” test is not very accurate. I’d go with likelihood ratio \chi^2 tests.

Thanks a lot for your valuable answer and I do apologize for delay in reply because I just saw your answer today. I am afraid that the sample size (141) isn’t enough for me to perform logistic regression and to control other variables, and I would need larger sample size. But I didn’t understand you well regarding making the comorbidity as the dependent variable? because I am comparing comorbidity with prognosis (or better to say the risk of being amputated), so, the comorbidity here is independent while prognosis is dependent. Thanks again.

1 Like