Statical analysis - correlation

Submitted by Patrizia on Tue, 05/29/2018 - 09:40

Dear Chao-Gan,

in SPM, when a correlation analysis in performed, it is necessary to specify the study design (eg. multiple regression).

I am wondering what is the model used in DPABI->Statistical Analysis->Correlation Analysis.

Thank you,

P.

 

Forums:

YAN Chao-Gan

Wed, 05/30/2018 - 06:11

It's simply Pearson correlation coefficient.

If you added covariates, then it's partial correlation.

Patrizia

Thu, 05/31/2018 - 09:02

Thank you for your answer. 

Yes, I have included two covariates.

Just to be clear, does the correlation implemented in DPARSFA use the GLM function of SPM?  

In any case, is specified any model? Clearly some modeling must have been done but the methods description does not make it apparent what it was.

 

Thanks.

YAN Chao-Gan

Fri, 06/08/2018 - 05:55

You can look at y_Correlation_Image.m

[b_OLS_brain, t_OLS_brain, TTest1_T, r_OLS_brain, Header] = y_GroupAnalysis_Image(DependentVolume,Regressors,OutputName,MaskFile,CovariateVolume,Contrast,'T',0,Header);

 

Df_E = size(Regressors,1) - size(Contrast,2);

    

rCorr = TTest1_T./(sqrt(Df_E+TTest1_T.*TTest1_T));

It used a regression model, and then convert the T value to r value.