# T-Test Analysis

#### BClemenson

##### New Member
Hello, for a course I am taking, I need to understand superficially an applied statistical approach in a research paper.
https://www.pnas.org/content/pnas/116/21/10250.full.pdf
The hyperlink above was the article I selected in which it dealt with a descriptive statistical approach of using peripheral blood mononuclear cell (PBMC) as a biomarker for the diagnosis of chronic fatigue syndrome. The analysis of 40 patients, 20 with CFS and 20 controlled, was done using a T-test that calculated the P-value of percent change in impedance between the plateau and both the baseline and minimum when the PBMC was exposed to a stressor. The P-value calculated were 7.72E-9 when looking at the percent change between the minimum to the plateau and a 4.48E-9 P-value for the percent change between baseline and plateau. The lower the P-value the more likely it is that you can reject the null hypothesis and the higher the value you would fail to reject the null hypothesis, and it would be more likely that there was not any difference between the values. In this case, the P-values are quite small indicating that there is a statistical difference. However, a statistical difference shouldn’t be enough to indicate if the data is applicable in a healthcare setting. Therefore, it would be important to look at the clinical values as well as more research before determining if it should be used as a diagnostic test for CFS. I wanted to make sure my understanding of the statistical approach is correct.
@hlsmith

#### hlsmith

##### Less is more. Stay pure. Stay poor.
@BClemenson - this is interesting. Was there a large difference reported between groups and did they place confidence intervals around the difference? Lastly, rhetorically, what value should be excluded from the confidence interval for a difference? I will note, not having looked at the article, with a sample of only around 40 subjects and such a small pvalue - one would have to imagine the difference between groups most be fairly large. Also the standard deviations (dispersion) around the means must not be too large, because if they were, it becomes difficult to say the two mean values may not be different.