Last Thursday you talked about the large number of features needed to ensure capturing all the characteristics for each graph (2^(n^2)) when applying statistical theory on graphs. So, as you told us, we have to choose a few features to compute and apply the statistical theory on them. I know by that we have discarded a lot of the features of the graphs, but is it necessary to test for all features?
For example, if you choose a few features to observe and find that there is a significant difference between two groups of people. Is it not sufficient to say that you can distinguish between these two groups using these specific features only?
Last Thursday you talked about the large number of features needed to ensure capturing all the characteristics for each graph (2^(n^2)) when applying statistical theory on graphs. So, as you told us, we have to choose a few features to compute and apply the statistical theory on them. I know by that we have discarded a lot of the features of the graphs, but is it necessary to test for all features? For example, if you choose a few features to observe and find that there is a significant difference between two groups of people. Is it not sufficient to say that you can distinguish between these two groups using these specific features only?