Dear experts,

I received all NBS output and input files from a previous analysis. Significant changes among the three groups were only reported with a figure (see attached). In addition to this figure, I would like to report edge weights with a table as well. Where in the NBS structure can I find the connectivity (correlation) matrix for each of the three groups? Further, how can I obtain p value for each significant connection?

Many thanks for your help and best wishes,
André

Hi Andre

You should be able to get the edge weights from the original matrices that you provided as input to the NBS. You could use nbs.NBS.con_mat{1} as the mask to identify the relevant connections in the original connectivity matrices. 

p-values for each individual connection are not provided. In general, it is not valid to make inference and conclusions about individual connections with the NBS. However, if you really want, you could convert the test statistics for each edge stored in nbs.NBS.test_stats into p-values.

Page 31 of the reference manual has further details about the variables stored in the NBS structure.

I hope that helps.

ANdrew


Originally posted by André Schmidt:

Hi Andrew,

thank you so much for your time and help, very much appreciated.

I found the individual matrices in the input folder. As I would like report edge weights for each group, is it OK to simply average multiple individual matrices by element?

In a second step, I would like to present statistics for each of the three group comparisons, in support of the attached figure. I think nbs.NBS.test_stat provides T values for each edge, right? Can you please tell me how I can see which of the edges were significantly different across groups? I could then highlight the relevant T values in nbs.NBS.test_stat.
You mentioned to use nbs.NBS.con_mat{1} to identify significant connections, but how exactly can I do this?

Many thanks again for your kind help, and my apologies for this trivial questions,
André

Hi Andre,

Yes - I think that averaging edge weights across the set of significant edges for each of the three groups sounds very reasonable to me. 

Yes - you are right. nbs.NBS.test_stat provides the T value for each edge.

Here is some code to get the T values for the significant connections. It will print the T values to the screen

ind=find(nbs.NBS.con_mat{1}); %find the index of the significant edges
x=nbs.NBS.test_stat;    %store the T values in a matrix called x
x(ind)     %print to the screen the T values for the significant connections.

Andrew


Originally posted by André Schmidt: