I am currently trying to do Fisher's exact tests on SPSS V20 as I had been advised this is correct test to undertake. What I am trying to do is to compare medical results achieved in a small cohort of local patients against data from clinical trials for 3 different drugs i.e. compare the percentage of patients who achieved e.g. undetectable viral load at 48 weeks of treatment for each of the 3 drugs.
This I understand to be categorical data. Within each data group the number of patients varies from 4 to 15 in my data compared to 250-350 in the clinical trials.
Does a Fisher's exact test sound correct? Can you advise how I can obtain the 95% confidence interval?
Any advice would be hugely appreciated!
Last edited by bowerpower83; 05-30-2012 at 03:48 PM.
I have tried to make my thoughts a little clearer but I'm still not sure about what I should be doing.
What I am trying to do is to compare results (yes/no
response) following drug treatment for my group of patients against
the same results from a clinical trial using a different cohort of
patients i.e. percentage achieved x. I understand that this data is
independent and categorical.
From my cohort of patients each data group contains n=4-15 compared to
in n=250-350 the clinical trial.
I thought that Fishers Exact test was the appropriate test since
patient numbers are small however I am unsure whether I should be
actually be doing a Z test.
Yes that is correct. Since some of my data/results only includes 4 or 6 patients so am I right in thinking a Fishers Exact is more appropriate than Chi Squared?
A colleague suggested that I should be performing a Z test (not Fishers) which apparently answers the question: are the proportions of A the same in two cohorts i.e. my patients v clinical trial patients. He says that Fishers Exact test should be limited to data within one study e.g. data from my patients only because Fishers answers the question - is there an association between variable A and variable B.
Can you possible shed any light on this as I am quite confused?