Iterate thru each row on a Matrix and return the result on a vector

Hello, this is the first question I'm asking here, after a couple of reading either in Matlab help, or the forum.

I haven't been able to get what I'm looking for.

So, I'm working with a set x = 213 x 9

I'd like to perfrom a function for each row, therefore it should be a vector of 213 x .

After that plot the final result vs the original data.

Example of DataSet

See attached image

Function to be performed

function KU =  Kurtosis( y )
% % Kurtosis 
    T=size(y,1);
    KU(1)=y(1);
    for i=2:T
      y_mean=mean(y(1:i));
      y_std=std(y(1:i));
      KU(i)=sum(((y(1:i))-repmat(y_mean,i,1)).^4)/((T-1)*y_std.^4);
    end
end

I've tried the following:

Using for loop

Kurtosis_test = 0;
for i = 1:size(data(1:1,:))
    for j = 1:size(data)
    Kurtosis_test = Kurtosis(data(i:j,:))
    end
end

Using rowfun

test = rowfun(Kurtosis,data)
Not enough input arguments.

Error in Kurtosis (line 3)
    T=size(y,1);

Thanks in advance.

Answer 1

[dpb]

Don't need no steenkin' loops! This is MATLAB!!! :)

kurt=kurtosis(y.').';

Above uses the fact that most ML functions in general and kurtosis in particular are written to operate on matrices by column; hence the ".'" to transpose your row-oriented array to columns first, then to transform the result back to a column to match rows.

mean, std and friends all operate same way.

NB: all of these also have an optional DIM argument that can be used to avoid the explicit transpose by saying which dimension of the array to operate over. In this case you want to operate on rows which is second dimension so

kurt=kurtosis(y,[],2);   % operate by rows, not default column

NB2: the [] placeholder, the DIM argument is the third, the second optional argument is a flag for bias correction (default is on). See

doc kurtosis

for details of this.

BTW, the answer to the original question because every once in a while it is necessary to loop over an array, would be something like

nr=size(y,1);                % number rows in array
kurt=zeros(nr,1);            % preallocate for result
for i=1:nr
  kurt(i)=kurtosis(y(i,:));  % compute kurtosis over each row vector in turn
end

Obviously, the function called can be any that accepts a vector, not just kurtosis, I just used for simplicity to illustrate the calling syntax.

[ALBERTO García Anaya]

Thanks a lot ! You have helped me a lot !