I have data in the form of percentages of patients in Cohort 1 and Cohort 2 that display the clinical presentation of certain symptoms including splenomegaly, short stature, history of failure to thrive, lymphadenopathy, jaundice among others. I am trying to choose what statistical test would be appropriate to compare percentage of patients in the two cohorts that present in clinical with a certain symptom while also knowing the sample size.
I use the stat package below (online and free) and they suggested a two sample T test for percentages but I am getting different opinions from hospital statisticians.
https://www.statpac.com/statistics-calculator/percents.htm
Use t test for testing the significance of difference between the two cohorts.
F test is also suitable/applicable in this case.
You have two groups, then for comparing of means, if their distribution are normal, you can use T-test other with you can transform the data to be normally distributed and use T test
If the data still free distribution (other than normal distribution) you have to use non parametric test as Mann–Whitney
I do agree with Khaled El-khawas.
I like to add one more option.
Since the objective is to analyze the percentages of patients in 2 groups namely Cohort 1 and Cohort 2, the test of significance of difference between the proportions of two groups (binomial test) can also be applicable.
If you are comparing difference in proportion between two cohorts in symptoms and other clinical features use chisquare test or fishers exact tests.
In order to apply Chi-square test, as mentioned by Ravishankar Suryanarayana, the two distributions of the percentages of patients in 2 groups namely Cohort 1 and Cohort 2 are required to be converted in to two frequency distributions.