Storing index logic in new column per iteration

For each idx variable, I want the values to be stored in a new column per iteration. Right now, each idx is being overwritten per loop iteration. I do not want the idx logic to be stored in one column with new rows; each iteration should make a new column in the idx logic field.

for j = 2:length(fieldnames(data))

for i = 1:(length(fieldnames(data))-1)

firstoutput = data.(sprintf("variables_%d", i));

secondoutput = data.(sprintf("variables_%d",j));

data.idx2 = ismember(firstoutput(:,2),secondoutput(:,2));

data.idx3 = ismember(firstoutput(:,3),secondoutput(:,3));

data.idx4 = ismember(firstoutput(:,4),secondoutput(:,4));

end

end

Answer 1

[Timothy]

You would need to use i and/or j, depending on how you are thinking of an "iteration" for your algorithm, as indices into your data.idx# fields. If i and/or j are large, I would also recommend pre-allocating them as a logical/boolean vector/matrix, as appropriate for your definition of iteration.

For example, it seems like you would want to store all combinations of i and j, based on your code so you might do something like this:

L = length(fieldnames(data));
data.idx2 = zeros(L-1, L-1, 'logical');
data.idx3 = data.idx2;
data.idx4 = data.idx2;

for j = 2:L
 for i = 1:L-1
  firstoutput = data.(sprintf("variables_%d", i));
  secondoutput = data.(sprintf("variables_%d",j));
  data.idx2(i,j) = ismember(firstoutput(:,2),secondoutput(:,2));
  data.idx3(i,j) = ismember(firstoutput(:,3),secondoutput(:,3));
  data.idx4(i,j) = ismember(firstoutput(:,4),secondoutput(:,4));
 end
end

[Annette Ciecierega]

I get the following error when I try your suggestion:

"Unable to perform assignment because the size of the left side is 1-by-1 and the size of the right side is 274-by-1."

Which has been the error that I have been struggling with

[Timothy]

On which line do you get this error?

[Annette Ciecierega]

data.idx2(i,j) = ismember(firstoutput(:,2),secondoutput(:,2));

here is what I wrote from your suggestions

L = length(fieldnames(data));

data.idx2 = zeros(L-1, L-1, 'logical');

data.idx3 = data.idx2;

data.idx4 = data.idx2;

for j = 2:L

for i = 1:L-1

firstoutput = data.(sprintf("variables_%d", i));

secondoutput = data.(sprintf("variables_%d",j));

data.idx2(i,j) = ismember(firstoutput(:,2),secondoutput(:,2));

data.idx3(i,j) = ismember(firstoutput(:,3),secondoutput(:,3));

data.idx4(i,j) = ismember(firstoutput(:,4),secondoutput(:,4));

end

end

[Timothy]

Ah, OK, I see what I missed now that I'm looking at it again. The output of is member is different than what I was initially thinking. You just need another dimension on your data.idx# fields to accomodate the output of ismember. Is the length of data.variables_## fixed or variable? If it is fixed, you can add the third dimension size to the zeros call to pre-allocate your data.idx# fields. If it is variable, then you may want to make your data.idx# fields into any array of cells instead.

For cell arrays:

Preallocate using ... = cell(L-1);

Access using {} instead of () for data.idx#, like data.idx3{i,j} = ...

[Annette Ciecierega]

Thank you for your help.

[Timothy]

You're welcome, were you able to get something working the way you wanted?

[Annette Ciecierega]

I actually posted another question:

https://www.mathworks.com/matlabcentral/answers/1578430-plotting-repeating-values-in-excel-files-vs-the-files-that-they-overlap-in