help
86 Subscribers 
help > binary matrices
Showing 1-5 of 5 posts
Sep 29, 2015 12:09 PM | Peter McColgan
binary matrices
Dear Andrew,
I am investigating group differences in binary matrices and get a large test statistic of 11.09 for 5 connections in the network, I use a contrast of -1,1,0,0,0 which includes 3 convariates.
Do this sound reasonable, I wanted to check as the t-stat is so high ?
Thanks
Peter
I am investigating group differences in binary matrices and get a large test statistic of 11.09 for 5 connections in the network, I use a contrast of -1,1,0,0,0 which includes 3 convariates.
Do this sound reasonable, I wanted to check as the t-stat is so high ?
Thanks
Peter
Sep 30, 2015 01:09 AM | Andrew Zalesky
RE: binary matrices
Originally posted by Peter McColgan:
Dear Andrew,
I am investigating group differences in binary matrices and get a large test statistic of 11.09 for 5 connections in the network, I use a contrast of -1,1,0,0,0 which includes 3 convariates.
Do this sound reasonable, I wanted to check as the t-stat is so high ?
Thanks
Peter
I am investigating group differences in binary matrices and get a large test statistic of 11.09 for 5 connections in the network, I use a contrast of -1,1,0,0,0 which includes 3 convariates.
Do this sound reasonable, I wanted to check as the t-stat is so high ?
Thanks
Peter
Sep 30, 2015 01:09 AM | Andrew Zalesky
RE: binary matrices
Hi Peter,
Have you tried using weighted matrices?
The matrices should not really be binarized before NBS, since we are interested in differences in edge strength, which are lost after binarization.
Andrew
Originally posted by Peter McColgan:
Have you tried using weighted matrices?
The matrices should not really be binarized before NBS, since we are interested in differences in edge strength, which are lost after binarization.
Andrew
Originally posted by Peter McColgan:
Dear Andrew,
I am investigating group differences in binary matrices and get a large test statistic of 11.09 for 5 connections in the network, I use a contrast of -1,1,0,0,0 which includes 3 convariates.
Do this sound reasonable, I wanted to check as the t-stat is so high ?
Thanks
Peter
I am investigating group differences in binary matrices and get a large test statistic of 11.09 for 5 connections in the network, I use a contrast of -1,1,0,0,0 which includes 3 convariates.
Do this sound reasonable, I wanted to check as the t-stat is so high ?
Thanks
Peter
Sep 30, 2015 05:09 AM | Peter McColgan
RE: binary matrices
Thanks Andrew, I was interested to look specifically at binarised
matrices only, is it valid / possible to use them in NBS ?
In using them for the rest of my analysis I get very coherent and biologically plausible results (particularly with clinical correlations).
bw
Peter
In using them for the rest of my analysis I get very coherent and biologically plausible results (particularly with clinical correlations).
bw
Peter
Sep 30, 2015 05:09 AM | Andrew Zalesky
RE: binary matrices
Hi Peter,
Using binary matrices is feasible, although this eliminates the information contained in the connectivity weights.
Glad to hear that your results are coherent.
Andrew
Originally posted by Peter McColgan:
Using binary matrices is feasible, although this eliminates the information contained in the connectivity weights.
Glad to hear that your results are coherent.
Andrew
Originally posted by Peter McColgan:
Thanks Andrew, I was interested to look
specifically at binarised matrices only, is it valid / possible to
use them in NBS ?
In using them for the rest of my analysis I get very coherent and biologically plausible results (particularly with clinical correlations).
bw
Peter
In using them for the rest of my analysis I get very coherent and biologically plausible results (particularly with clinical correlations).
bw
Peter