# Subtract mean from each table by columns

25 views (last 30 days)
Commented: Image Analyst on 22 Jun 2020
Hello,
Assume we have the following table
t = array2table([rand(10,1),rand(10,1),rand(10,1)],'VariableNames',{'feat_1', 'feat_2','feat_3'});
t =
10×3 table
feat_1 feat_2 feat_3
________ ________ _________
0.12991 0.60198 0.82582
0.56882 0.26297 0.53834
0.46939 0.65408 0.99613
0.011902 0.68921 0.078176
0.33712 0.74815 0.44268
0.16218 0.45054 0.10665
0.79428 0.083821 0.9619
0.31122 0.22898 0.0046342
0.52853 0.91334 0.77491
0.16565 0.15238 0.8173
I can get the mean and standard deviation for each solumn using
meanArray = mean(table2array(t,1));
stdArray = std(table2array(t,1));
meanArray =
0.3479 0.4785 0.5547
stdArray =
0.2415 0.2836 0.3793
Is there any vector way to caclulate (element - mean / std) on every element in table without loop and keep the table? Thanks
I know normalize could standardize data but I need to make it in code.

#### 1 Comment

madhan ravi on 14 Jun 2020
Expected output?

Image Analyst on 14 Jun 2020
Edited: Image Analyst on 14 Jun 2020
Try this:
t = array2table([rand(10,1),rand(10,1),rand(10,1)],'VariableNames',{'feat_1', 'feat_2','feat_3'})
m = table2array(t);
meanArray = mean(m)
stdArray = std(m)
z = (m - meanArray) ./ stdArray % a matrix
zt = array2table(z, 'VariableNames',{'feat_1', 'feat_2','feat_3'}) % a table

#### 1 Comment

Thank you so much.

Hello,
I came by another solution using bsxfun
Here is a complete example based on my question
% Generate random columns for the example (10 by 3)
t = array2table([rand(10,1),rand(10,1),rand(10,1)],'VariableNames',...
{'feat_1', 'feat_2','feat_3'});
% Subtract from the mean vector (3 by 1)
t1 = bsxfun(@minus, mean(table2array(t)),table2array(t));
% Divide by stadard deviation vector (3 by 1) using right array division
t2 = bsxfun(@rdivide,t1,std(table2array(t)));
% Create the standardized table
t = array2table(t2,'VariableNames',{'feat_1', 'feat_2','feat_3'});
% Check mean and std of feat_1
feat_1_mu = mean(t.feat_1)
feat_1_sig = std(t.feat_1)
feat_1_mu =
0
feat_1_sig =
1

#### 1 Comment

Image Analyst on 22 Jun 2020
Yeah, you can do it that way if you want. I always find their *xfun() functions very cryptic and confusing than the regular math symbols. Plus you always have to find or figure out what the operations are. I mean, who would know, off the top of their head, that if you want to divide something you use @rdivide instead of the slash symbol /, or @minus instead of -. And it gets even more confusing for other operations.

R2020a

### Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!