yalemrrc/cpm

cpm's Issues

Error: Not enough points to perform robust estimation.

Greetings,

First and foremost, thank you for this code. I am trying to use the CPM with my data (124 subjects), using the following basic command:

[y_predict, performance]= cpm_main(curdata, behvofi, 'pthresh',0.01,'kfolds',10)

I get no errors when using a p-threshold of 0.05, but as soon as I use 0.01 or smaller, I run into the following error.

# Running over 10 Folds.
Performing fold no. 1 2 Error using statrobustfit (line 21)
Not enough points to perform robust estimation.

Error in robustfit (line 114)
[varargout{:}] = statrobustfit(X,y,wfun,tune,wasnan,doconst,priorw,dowarn);

Error in cpm_train (line 22)
mdl=robustfit(summary_feature,y');

Error in cpm_cv (line 38)
    [~, ~, pmask, mdl] = cpm_train(x_train, y_train,pthresh);

Error in cpm_main (line 53)
[y_predict]=cpm_cv(x,y,pthresh,kfolds);

I am just curious as to what is the nature of this error? It is suggesting that there are not enough significant edges (at the p-threshold) to apply the CPM?

Any insight would be greatly appreciated.

Thank you,
Paul

which one is the rcpm

I see serveral rCPM code,so which one should I use to conduct a rCPM?

python code question

Hello,

Thanks for the great resource! I am trying to run CPM with the python code. In ln 17 of train_cpm I get the following error:
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

Any ideas? Thanks!

Best,
Alex

CPM

Hello, thanks so much for the CPM toolbox. Its a very resourceful. My question is do you suggest that behavioral scores should be normalized while pairing them with the connectivity matrix?

I cannot find the code for cCPM

Thank you for providing the useful code. The paper (Gao et al., 2019) mentions that the script for CCA based CPM is provided here, but I cannot find the cCPM script. Could you guide me where it is? Thank you again.

Kind regards,
Yu-Chi Chen

How to deal with NAs in connectivity matrices

First of all, thanks so much for sharing your code with the community.
My dataset has some NAs because we exclude voxels with low SNR and some ROIs had no voxels survive.
Is this a problem? The code seems to run fine (I get an NA warning but also get the results).

Any advice appreciated!
Maria

Adding covariates

Hi, I am interested in using this toolbox to associate biomarkers with structural connectivity matrices. Could covariates be included in the model? If not, what recommendation would you give me? Adjusting the biomarker scores with a linear regression for the covariates could be an option?

Greetings and thanks in advance

Dimensions of input connectivity matrix

Thank you all for this extremely valuable code.

Apologizes if this question should not be posted here.

My question is: what format should the input matrix be exactly? I have 120 participants, each with their own single connectivity matrix (148 x 148). Should the input matrix be than 148 x 148 x 120? Or should the connectivity matrix be turned into a vector before running the code, so in my case, the dimensions would a two dimension matrix with length as (148 x 148) and rows being 120?

Thank you in advance.

I can not find the 'predict_behavior.m' script

Thank you for providing the useful code. I'm trying to reproduce the code form the paper 'Using connectome-based predictive modeling to predict individual behavior from brain connectivity' , but I cannot find the 'predict_behavior.m' script mentioned in the 'calculate the true predict correlation'part. Could you guide me where it is? Thank you again.

Kind regards,
Xuan Liu

Sample data

Hello,
Thank you for sharing your CPM code.
I did some tests with my own data, but the CPM did not seem to perform well as expected. I was wondering would it be possible to share some sample data (e.g., FC matrix and behavioral variable) on which CPM performed well?