How do I go about this? And what statistical test can I use?

Question

Hi,

I’m currently doing a science project on sickle cell disease, I’m researching to see how a well a few drugs can induce fetal haemoglobin in sickle cell disease patients.

Currently what I’ve done is gather results from clinical studies, in which I took the mean levels of fetal haemoglobin in all patients before the trial and the means of the level of fetal haemoglobin after the trial. I then put the numbers in a T test to calculate the efficiency of the drug to induce fetal haemoglobin. Is this correct?

If not how can I go about this? And what test can I use, id appreciate all your help. Thank you

Answer 1

This is correct. If you only have one thing that you are testing against (haemoglobin) then you do this. However, you need to be sure that you use the right T-test. The t.test function in excel will do this. One question you have to ask is how to set up your null and alternative hypothesis (mostly the alternative).

The null hypothesis would most likely be that the two means are the same (mu1=mu2).

Now you have options for the alternative hypothesis. If you are trying to measure if they are different (Not sure which one has the disease) you can use a two-tailed test (mu1/=mu2). If you are sure that one direction indicates the disease, you can do a one-tailed test either left or right (looking for high or low numbers). This is done with mu1<mu2 or mu1>mu2.

In excel:

t.test(array1, array2, tail, type)

array1=data for mu1

array2=data for mu2

tail=type of tail, 1=one tailed, 2=two tailed

type=type of test 1=paired (unless you have paired, don't choose) 2=equal variance (you have the same variance or sd for the tests. If you getting data from other studies likely not). 3=unequal variance.

Then you see if the p value that you get is below your threshold. This should be set depending on the field. In general, 0.05 is default. However, with medicine it can be 0.01 or 0.001. I would check with whoever is running the class to be sure. Either way, if the p value that the t.test gives is below that value, you can reject the null hypothesis, leaving the alternative. If it does not, you fail to reject the null hypothesis!