Take samples from the previous sample

Hi all, I have set of data, the size doesn't really matter. Lets call it x1. I want to take a random set of values from x1 and lets call it x2. Then take random samples from x2 and call it x3. So on and so forth. I am not really sure how to do this, I have tried a couple of things, but without luck. Any help is appreciated. Also, I have access to the statistics library. The things that I have tried made use of the function randsample.

Thanks.

[Azzi Abdelmalek]

Can you give more details? how many sample? when do you want to stop?

[Ellie]

It doesn't really matter, but for arguments sake: sample size 100, and stop at 10.

Answer 1

[Star Strider]

I’m not certain what you want, but this is one possibility:

x1 = randsample(99, 15)'
x2 = x1(randsample(15,10))

x1 =

      76    68    83    17    87    16    35    30    53    62    26    38     8    98    40

x2 =

       8    53    40    38    17    83    26    62    35    30

The only constraint is to be certain that in ‘x2’ and beyond, the first argument to randsample is not greater than the length of the vector created in ‘x1’ or the earlier vectors.

[Ellie]

That is the desired output, but how could I make it more general. For example, if I want to do that 10 times, I don't want to have to do that over and over.

[Star Strider]

One option is to put it into a loop:

x{1} = randsample(99, 15)';
for k1 = 2:10
    x{k1} = randsample(x{k1-1},15);
end

All the vectors are the same sizes, but since they are cells, you can experiment with them to get the result you want.

EDIT — To use randsample with replacement, include the third argument as true or 1.

Answer 2

[Azzi Abdelmalek]

x=1:100
k=1;
out{k}=x
while numel(out{k})>10
  n=numel(out{k});
  idx=randperm(n-1);
  out{k+1}=out{k}(idx);
  k=k+1;
end
celldisp(out)

[Ellie]

Is there a way to do this with replacement?

And maybe without using cells?

[Azzi Abdelmalek]

the new samples have different sizes, you can't avoid cells. Why do you want to avoid cells?