How to run a paired samples t-test?

Hello, 

 

I am trying to conduct a paired samples t-test, looking at differences in graph metrics pre and post an intervention. However, I am having some difficulty setting up this analysis in GraphVar.

 

I followed the within design I tutorial, however that is looking at change in a graph metric i.e. efficiency in T1 and T2 between sex. However, I am only interested in changes in graph metrics between T1 (pre) and T2 (post).

 

I have attached my variable sheet to show what I have currently been doing. For the GLM, I selected group (pre and post) as the between factor and subj_ID as the Within ID. However, the results I get do not look right (see attached Results). The yellow line with filled in circles is telling me that the t-test is significant. However, the two lines representing pre and post values are superimposed on each other and at zero.

 

How would I set up my variable sheet as well as the GLM to run a paired samples t-test?

 

Thank you in advance. 

AttachmentSize
Image icon Results .png107.56 KB
File Variables.xlsx9.03 KB

Forums:

Hi,
your variable sheet is correct. However, you have to select "group" (pre post) as "Within Covariate" and NOT "Between factor". It is correct that you selected subj_ID as Within ID.
GraphVar will subsequently subtract the pre and post values per individual and performs a one-sample t-test agains zero (which is essentially a paired sample t-test). Therefore you will not have the possibility to show both groups (pre versus post) in the viewer but will only see one line (for your global clustering) that represents the difference of pre vs post.

Thank you for your response. 

Given that only one line is generated, if we find that there is a significant difference pre and post intervention, how would we interpret the direction of the finding? For example, if there is a significant difference in global clustering coefficient, how do we know if the global clustering coefficient is larger or smaller after the intervention? 

Is the answer in the mat file (1 row x 26 columns) created at each threshold for a particular graph metric? We have 13 subjects with pre and post data. Does each column represent a subjects pre and post values? Further, how is this data organized? Are the first 13 columns all pre values and the last 13 columns all post values?

Thank you in advance.

 

Hi and sorry for the very late reply,
GraphVar subtracts the first group from the second group (e.g. first group = pre, second group = post; post-pre = diff). Yes, the answer is in the mat file. Here the columns represent the subjects and this is ordered exactly as the subjects in your variable sheet. Again, I am sorry for the late reply!
Johann

No worries - that makes sense!

Thanks Johann!