A question for processing data

[Joey Gao]

Dear professors,

I noticed that the range of my processed data from the tvbpipeline is (0, 100, 569628). While most pictures of papers or jupyternotebook tutorials, the bar is 0.5 or 0.005 differently. And in tvb-dataset the weights of any data is from zero to 2. 

So is there any special processing to process the SC metric? I think the most parameter of tckgen is larger than 1M. And I noticed that after tckgen processing there is a subsample process to 100K in pipeline.

So after I get the sc metric from pipeline with (0, 100, 569628). Just normalized it to 0 to 1 is right?

Thanks very much. 

[Michael Schirner (michael.schirner@bih-charite.de)]

Dear Joey,

yes, many researchers normalize the SC after tractography. For example, by dividing every element by the 95th percentile of all SC values over the entire cohort of SCs (being careful to not simply divide each SC for each subject by its specific maximum or N-th percentile value, but to use one value globally for the entire cohort, to not remove effects that could explain differences between subjects). To get from these relative connection weights to absolute weights in terms of a brain network model, the SC is then multiplied in TVB by a global coupling scaling factor, which rescales the entire SC. This is often one of the first parameters that is fit, when an SC is used for a simulation.

Best,

Micha

[Joey Gao]

Thanks again. I learned a lot from that.