Problems with group comparisons using non parametric statistics

2 次查看（过去 30 天）

显示更早的评论

Eric 2014-9-22

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/155839-problems-with-group-comparisons-using-non-parametric-statistics

关闭： MATLAB Answer Bot 2021-8-20

在 MATLAB Online 中打开

Hi all,

I have used EEGLABs function 'spectopo' to extract the power values and later found the band specific values. Now, I'm interested in using nonparametric statistics to test for significance.

What I'm trying to do is to divide the sets (15 sets x 19 channels) in to three groups and do the comparisons between the mean value of the power in each electrode like group 1 vs group 2, group 2 vs group 3 and so on. Afterwards, I'm trying to find the significant p-values. The only problem is that I can't get it to work probably because it either saying that the index exceeds the matrix dimentions or that it compares the values for 1 electrode (p-values are then 1).

I hope that someone can help.

Script:

load('Powerinalpha.mat');

for freqband = 1:2 switch freqb pow = Alpha; pow = Beta; end

pvalue = 0.05;

    for session = 1:2     
           matr = [];
           if session==1
               matr = pow(1:5,:,:);
           else
               matr = dataset(6:10,:,:);
           end
           for channel = 1:21
                Pre = datamat(:,:,channel);
                Post = datamat(:,:,channel);
                [p, h, stats] = signrank(squeeze(Post), squeeze(Pre), 'alpha', pcrit);
                Values.sig(freqb,session,channel) = h;
                Values.pval(freqb,session,channel) = p;
                Values.signedrank(freqb,session,channel) = stats.signedrank;
           end
           nonpara.channelsig{freqband,session} = find(Values.pval(freqband,session,:)<pcrit);
        end
end