Help! Which test should I use?

mob

New Member
#1
Im conducting a study in which there are 2 groups of uneven size (A=55, B=65).
I'd like to compare the amount of times each patient was examined - what's the appropriate test?
I'd also like to compare the amount of times/per hour each patient was examined, is that a different test now?
 
#2
Im thinkin it is probably better to just do times/per hour and total hours of examination. The former is probably best to just log transform and t-test, (assuming you do not got to much zeros or too close to it). The latter, probably the same. usual t-test caveats apply.
 

noetsi

Fortran must die
#3
What is your dependent variable, how is it measured. You have not suggested what you want to test. You can just show the mean time patients were examined if there is nothing to test.

What is your hypothesis.
 

mob

New Member
#4
Thank you all for replying.
The null hypothesis is that the patients were test the same amount of times in both groups and I'd like to reject that hypothsis.
The problem is that the patients were not admitted for even amounts of time and if a patient is admitted for a longer time he will be examined more times.
I thought I'd compare the median amount of examinations- Mann Whitney?
and compare the mean amount/hour (could a T-test be appropriate? Just because opposed to amount of exams time is a contious variable)
 

noetsi

Fortran must die
#6
I don't understand the log transform question. There are commonly problems in the data not having a normal distribution that leads you to log it.

I am not an expert in what you asking. I think you could add a mediating (another predictor) variable, time in the system, to control for this, but I am not sure that would work in this case. This is where having theory, past reports, would really help.