help > ROC curve analysis
Showing 1-7 of 7 posts
replyDisplay: Results per page:
Nov 26, 2017  06:11 AM | Hua Yu
ROC curve analysis
Dear experts,

Recently, I made a two sample t test to compare the global functional connectivity between patients' group and HC group, and I did find some group differences. Now I want to see if the group FC connectivity differences could clarify the two groups , and what I supposed to do was to extract the FC value to do some simple ROC curve analysis in SPSS. Do you have suggestion about this?

Looking forward for your reply,

Many thanks,

Hua
Nov 26, 2017  02:11 PM | Andrew Zalesky
RE: ROC curve analysis
Hi Hua,

The NBS manual provides detailed instructions about how to extract FC values from significant subnetworks. It can be downloaded from NITRC.

Andrew

Originally posted by Hua Yu:
Dear experts,

Recently, I made a two sample t test to compare the global functional connectivity between patients' group and HC group, and I did find some group differences. Now I want to see if the group FC connectivity differences could clarify the two groups , and what I supposed to do was to extract the FC value to do some simple ROC curve analysis in SPSS. Do you have suggestion about this?

Looking forward for your reply,

Many thanks,

Hua
Nov 27, 2017  08:11 AM | Hua Yu
RE: ROC curve analysis
Dear Andrew,

Now I had the nbs.NBS con_mat file, and it stores the nodes information which showed significant group difference. The results comprised 1 significant network with 8 connctions. The binary connectivity val =
(12,16) 1
(4,32) 1
(9,46) 1
(12,46) 1
(32,46) 1
(9,128) 1
(16,144) 1
(19,144) 1

For the connectivity strength which I want to extract for each connection, how should I set the value of ' i & j', for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16? However,  this seems not right.

i=1; %Specify node 1 here
j=2; %Specify node 2 here
global nbs; N=nbs.STATS.N; ind_upper=find(triu(ones(N,N),1)); cross_ref=zeros(N,N); cross_ref(ind_upper)=1:length(ind_upper); cross_ref=cross_ref+cross_ref'; ind=cross_ref(i,j); fprintf('%0.2f\n',nbs.GLM.y(:,ind));

Can you help me modify the code or explain a little bit more about it?

Looking forward for you reply,

Sincerely,

Hua
Nov 27, 2017  03:11 PM | Andrew Zalesky
RE: ROC curve analysis
Hi Hua Yu,

for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16?

YES THIS IS CORRECT.

Andrew


Originally posted by Hua Yu:
Dear Andrew,

Now I had the nbs.NBS con_mat file, and it stores the nodes information which showed significant group difference. The results comprised 1 significant network with 8 connctions. The binary connectivity val =
(12,16) 1
(4,32) 1
(9,46) 1
(12,46) 1
(32,46) 1
(9,128) 1
(16,144) 1
(19,144) 1

For the connectivity strength which I want to extract for each connection, how should I set the value of ' i & j', for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16? However,  this seems not right.

i=1; %Specify node 1 here
j=2; %Specify node 2 here
global nbs; N=nbs.STATS.N; ind_upper=find(triu(ones(N,N),1)); cross_ref=zeros(N,N); cross_ref(ind_upper)=1:length(ind_upper); cross_ref=cross_ref+cross_ref'; ind=cross_ref(i,j); fprintf('%0.2f\n',nbs.GLM.y(:,ind));

Can you help me modify the code or explain a little bit more about it?

Looking forward for you reply,

Sincerely,

Hua
Nov 29, 2017  05:11 AM | Hua Yu
RE: ROC curve analysis
Dear Andrew,

Sorry to bother you so much. I think there's something still wrong with the code. First, let me make a brief introduction of my analysis. I divided my patient's and HCs' brain according to Dosenbach's  template in Dpabi which has 160 nodes and it generated 160x160 matrix for each subject. Now, I want to extract the connectivity coefficient for each pair of nodes that showed group connectivity difference which I uploaded before as you can see. However, every time when I type i=9, j=128 or i=14, j=144 or i=16, j=144 specifically for each pair of node number,  I would meet with the same problem showing that "Index exceeds matrix dimensions."

Can you help me solve this problem or do you need me to provide more information about my analysis?

Sincerely,

Hua
Originally posted by Andrew Zalesky:
Hi Hua Yu,

for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16?

YES THIS IS CORRECT.

Andrew


Originally posted by Hua Yu:
Dear Andrew,

Now I had the nbs.NBS con_mat file, and it stores the nodes information which showed significant group difference. The results comprised 1 significant network with 8 connctions. The binary connectivity val =
(12,16) 1
(4,32) 1
(9,46) 1
(12,46) 1
(32,46) 1
(9,128) 1
(16,144) 1
(19,144) 1

For the connectivity strength which I want to extract for each connection, how should I set the value of ' i & j', for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16? However,  this seems not right.

i=1; %Specify node 1 here
j=2; %Specify node 2 here
global nbs; N=nbs.STATS.N; ind_upper=find(triu(ones(N,N),1)); cross_ref=zeros(N,N); cross_ref(ind_upper)=1:length(ind_upper); cross_ref=cross_ref+cross_ref'; ind=cross_ref(i,j); fprintf('%0.2f\n',nbs.GLM.y(:,ind));

Can you help me modify the code or explain a little bit more about it?

Looking forward for you reply,

Sincerely,

Hua
Nov 29, 2017  04:11 PM | Andrew Zalesky
RE: ROC curve analysis
Hi Hua,

I'm not sure what might be causing this issue. You might want to seek assistance from a colleague with Matlab expertise.

Andrew
Originally posted by Hua Yu:
Dear Andrew,

Sorry to bother you so much. I think there's something still wrong with the code. First, let me make a brief introduction of my analysis. I divided my patient's and HCs' brain according to Dosenbach's  template in Dpabi which has 160 nodes and it generated 160x160 matrix for each subject. Now, I want to extract the connectivity coefficient for each pair of nodes that showed group connectivity difference which I uploaded before as you can see. However, every time when I type i=9, j=128 or i=14, j=144 or i=16, j=144 specifically for each pair of node number,  I would meet with the same problem showing that "Index exceeds matrix dimensions."

Can you help me solve this problem or do you need me to provide more information about my analysis?

Sincerely,

Hua
Originally posted by Andrew Zalesky:
Hi Hua Yu,

for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16?

YES THIS IS CORRECT.

Andrew


Originally posted by Hua Yu:
Dear Andrew,

Now I had the nbs.NBS con_mat file, and it stores the nodes information which showed significant group difference. The results comprised 1 significant network with 8 connctions. The binary connectivity val =
(12,16) 1
(4,32) 1
(9,46) 1
(12,46) 1
(32,46) 1
(9,128) 1
(16,144) 1
(19,144) 1

For the connectivity strength which I want to extract for each connection, how should I set the value of ' i & j', for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16? However,  this seems not right.

i=1; %Specify node 1 here
j=2; %Specify node 2 here
global nbs; N=nbs.STATS.N; ind_upper=find(triu(ones(N,N),1)); cross_ref=zeros(N,N); cross_ref(ind_upper)=1:length(ind_upper); cross_ref=cross_ref+cross_ref'; ind=cross_ref(i,j); fprintf('%0.2f\n',nbs.GLM.y(:,ind));

Can you help me modify the code or explain a little bit more about it?

Looking forward for you reply,

Sincerely,

Hua
Dec 1, 2017  03:12 AM | Hua Yu
RE: ROC curve analysis
Dear Andrew,

Thank you so much. Finally, I solved my problems and the code in the manual is totally right!

Many thanks again. You are so helpful.

HuaOriginally posted by Andrew Zalesky:
Hi Hua,

I'm not sure what might be causing this issue. You might want to seek assistance from a colleague with Matlab expertise.

Andrew
Originally posted by Hua Yu:
Dear Andrew,

Sorry to bother you so much. I think there's something still wrong with the code. First, let me make a brief introduction of my analysis. I divided my patient's and HCs' brain according to Dosenbach's  template in Dpabi which has 160 nodes and it generated 160x160 matrix for each subject. Now, I want to extract the connectivity coefficient for each pair of nodes that showed group connectivity difference which I uploaded before as you can see. However, every time when I type i=9, j=128 or i=14, j=144 or i=16, j=144 specifically for each pair of node number,  I would meet with the same problem showing that "Index exceeds matrix dimensions."

Can you help me solve this problem or do you need me to provide more information about my analysis?

Sincerely,

Hua
Originally posted by Andrew Zalesky:
Hi Hua Yu,

for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16?

YES THIS IS CORRECT.

Andrew


Originally posted by Hua Yu:
Dear Andrew,

Now I had the nbs.NBS con_mat file, and it stores the nodes information which showed significant group difference. The results comprised 1 significant network with 8 connctions. The binary connectivity val =
(12,16) 1
(4,32) 1
(9,46) 1
(12,46) 1
(32,46) 1
(9,128) 1
(16,144) 1
(19,144) 1

For the connectivity strength which I want to extract for each connection, how should I set the value of ' i & j', for example, if I want to extract the connectivity strength between node 12 and node 16, should I set the i=12, j=16? However,  this seems not right.

i=1; %Specify node 1 here
j=2; %Specify node 2 here
global nbs; N=nbs.STATS.N; ind_upper=find(triu(ones(N,N),1)); cross_ref=zeros(N,N); cross_ref(ind_upper)=1:length(ind_upper); cross_ref=cross_ref+cross_ref'; ind=cross_ref(i,j); fprintf('%0.2f\n',nbs.GLM.y(:,ind));

Can you help me modify the code or explain a little bit more about it?

Looking forward for you reply,

Sincerely,

Hua
reply