Hi Lizzie,

I suggest extracting the connectivity values for the connections comprising the significant components identified with the NBS. The NBS manual provides instructions on how to do this.

You could then show boxplots (between-group differences) or scatter plots (correlations) of the average connectivity across all connections comprising the component. Your plots could show confidence intervals, for example, to serve as an additional statistic.

You could also report an effect size. Once again, this can be achieved by averaging the test_stat values across the significant connections and dividing by the square root of the number of subjects, which yields Cohen's d (assuming that you are using a t-statistic). 

Andrew


Originally posted by Lizzie Shephard: