Matlab: Removing struct values via an index array - matlab

Question

I have a large struct (total) that I need to separate into 3 structs who's values are random selected from the original struct. I need a struct with 60% (trainingData), and two structs that are 20% each (testData & crossValData) but none of the values can overlap.
indexRand1 = zeros((size(total,1)*.2),1); % index of test data: 20%
indexRand2 = zeros((size(total,1)*.2),1); % index of cross validation data: 20%
for index_i = 1:size(indexRand1,1)
temp1 = randi([1 size(total,1)],1,1); % get a random value from total
temp2 = randi([1 size(total,1)],1,1); % get another random value from total
while ismember(temp1,indexRand1) % make sure 1st value is not already in test data
temp1 = randi([1 size(total,1)],1,1);
end
indexRand1(index_i,1) = temp1; % add 1st value to test data
while ismember(temp2,indexRand1) % make sure 2nd value is not already in test data
temp2 = randi([1 size(total,1)],1,1);
while ismember(temp2,indexRand2) % or cross validation data
temp2 = randi([1 size(total,1)],1,1);
end
end
indexRand2(index_i,1) = temp2; % add 2nd value to cross validation data
end
indexRand3 =[indexRand1;indexRand2]; % index of test and cross validation data
testData = total(indexRand1,:); % use index to get test data
crossValData = total(indexRand2,:); % use index to get cross validation data
total(indexRand3,1) = []; % remove test and cross validation data
trainingData = total; % save training data to new name
My problem comes at 'total(indexRand3,1) = []; % remove test and cross validation data' the error I get is 'A null assignment can have only one non-colon index.' How do I remove values from a struct using an index? (or how do you separate a struct randomly into 3 unequal structs?)

Answer 1

I think you are making this problem much harder than it needs to be. A more Matlab-y solution seems to be:
%Make some test data
% (Well, I guess you already have data. I need data to test with)
total(1000).sampleData = 1
%Determine the number of elements in each derived set
nTraining = round(numel(total)*0.6);
nTest = round(numel(total)*0.2);
nCrossVal = numel(total) - nTest - nTraining;
%Create a random order vector
ixsRandomOrder = randperm(numel(total));
%Use the random order vector to create distinct, derived sets
testData = total(ixsRandomOrder( (1:nTest) ));
trainingData = total(ixsRandomOrder(nTest + (1:nTraining) ));
crossValData = total(ixsRandomOrder(nTest + nTraining + (1:nCrossVal) ));