December 13, 2017 at 3:56 pm #21226
Hi BRAPH developers,
thank you for developing this great tool.
I just started using BRAPH and i have some perplexities. I tried to create a cohort (2 groups with one subject each) with two simple Undirected Binary Graphs (18 nodes each, 1 = edge between a pair of nodes 0 = no edge). The problem I encountered is that BRAPH reads the graphs as weighted, asking me to set a density/treshold in order to perform the analysis on binary graph.
Any insight on how to make BRAPH read my graphs as binary?
Thank you in advance,
GiulioDecember 14, 2017 at 10:02 pm #21229
If your matrices are already binarized (only contain ones and zeros) and you would like to include all the edges in your analyses, I believe that setting the density to 100% will allow you to do what you want.
If that doesn’t work, please send us one of your matrices and we will take a look.
JoanaDecember 15, 2017 at 9:54 am #21231
Thank you for your answer.
I tried setting various density and a strange thing happened: i set the graph density at 32% (roughly the actually density of my binary graphs, calculated by hand) and did some analysis (Nodal and Average Degree, Characteristic path lenght, Global efficiency and Cloustering coefficient). BRAPH calculated the global measures correctly (I had preaviously calculated them by hand), but gave me wrong Nodal measure (Nodal degree). Also, at 32% density, the matrices showed by BRAPH did not correspond to the ones I created.
I will send you the matricies I used if you want to investigate the matter futher. Anyway I was just using simple data wich I knew the structure off to familiarize with BRAPH: the data we will going to analize with the program will be weighted, so this will not be an issue.
Thank you again for your time.
GiulioDecember 20, 2017 at 12:01 pm #21240
The problem you are encountering when you enter binary graphs is due to the way the binarization/thresholding procedure is performed in Braph. Namely, when you enter a given density, a corresponding threshold is calculated according to which the matrix is binarized. In the case of binary graphs for which all edges are ones/zeros by construction, this may result in error since a given treshold cannot distinguish between the edges. Therefore, your matrices should correspond to the ones Braph creates at density of 100% (threshold 0) but in general might not correspond at different densities.
You are right, when you use weighted matrices this problem is automatically solved and you will be able to perform your analysis. However, if you could send me some of the matrices that you use, I would check this issue and maybe come up with a more detailed explanation.