Thread Subject:

Selecting subvectors and observations from original data to form samples

Hi!

I have a data, y=[379 357 390 376 376; 363 367 382 381 359; 401 402 407 402 396; 402 387 392 395 394; 415 405 396 390 395];
This has 5 rows(5 sub-vectors in my case) and 5 columns. I am supposed to draw 5 sub-vectors WITH REPLACEMENT from the original data vector and then draw 5 observations WITH REPLACEMENT from each sub-vector obtained. This leads to a new sample of size 5x5. Can anybody help me out with this using MATLAB? Am I supposed to use randsample or datasample?
Your help is much appreciated. Thank you.

Regards,
dhaskantha

"dhaskantha" wrote in message <jn5ea1$obe$1@newscl01ah.mathworks.com>...
> Hi!
>
> I have a data, y=[379 357 390 376 376; 363 367 382 381 359; 401 402 407 402 396; 402 387 392 395 394; 415 405 396 390 395];
> This has 5 rows(5 sub-vectors in my case) and 5 columns. I am supposed to draw 5 sub-vectors WITH REPLACEMENT from the original data vector and then draw 5 observations WITH REPLACEMENT from each sub-vector obtained. This leads to a new sample of size 5x5.

r = ceil(5*rand(1,5))
c = ceil(5*rand(1,5))
z = y(r,c);

% Bruno

"dhaskantha" wrote in message <jn5ea1$obe$1@newscl01ah.mathworks.com>...
> Hi!
>
> I have a data, y=[379 357 390 376 376; 363 367 382 381 359; 401 402 407 402 396; 402 387 392 395 394; 415 405 396 390 395];
> This has 5 rows(5 sub-vectors in my case) and 5 columns. I am supposed to draw 5 sub-vectors WITH REPLACEMENT from the original data vector and then draw 5 observations WITH REPLACEMENT from each sub-vector obtained. This leads to a new sample of size 5x5. Can anybody help me out with this using MATLAB? Am I supposed to use randsample or datasample?
> Your help is much appreciated. Thank you.
>
> Regards,
> dhaskantha
- - - - - - - - -
 p = ceil(5*rand(1,5));
 z = y(p,:);
 q = bsxfun(@plus,[1:5].',5*floor(5*rand(5,5)));
 z = reshape(z(q),5,5);

Roger Stafford

Obviously I and Roger read differently the problem:

- to me the observation-draw is the same for all sub-vectors,
- to Roger the observation-draw are different for each sub-vector.

Up to you to choose what you need.

Bruno

Many thanks Bruno and Roger. I will try this out.

Regards,
dhaskantha

Hi!

I am able to generate one sample with the way you proposed. I am supposed to generate N samples like z from the original data y. How do I use the loop to do it? Is there a better way for me to do so? Thank you.

Regards,
dhaskantha

> - - - - - - - - -
> p = ceil(5*rand(1,5));
> z = y(p,:);
> q = bsxfun(@plus,[1:5].',5*floor(5*rand(5,5)));
> z = reshape(z(q),5,5);
>
> Roger Stafford

"dhaskantha" wrote in message <jn848f$j9h$1@newscl01ah.mathworks.com>...
> I am able to generate one sample with the way you proposed. I am supposed to generate N samples like z from the original data y.
- - - - - - - - - -
 p = bsxfun(@plus,ceil(5*rand(5,1,N)),5*floor(5*rand(5,5,N)));
 z = reshape(y(p),5,5,N);

This should generate a three-dimensional array z of size 5 x 5 x N with elements from y. Each of the slices z(:,:,k) from k = 1 to k = N should be a valid "sample" from y in accordance with the random selection procedure you described.

  (This assumes that my interpretation of what you wrote is correct. Bruno had a different interpretation of your words, so you should decide which is the one you want to use.)

Roger Stafford

Hi Roger!

Thank you for your help. Will try working out using this method.

Regards,
dhaskantha


"Roger Stafford" wrote in message <jn9msu$nri$1@newscl01ah.mathworks.com>...
> "dhaskantha" wrote in message <jn848f$j9h$1@newscl01ah.mathworks.com>...
> > I am able to generate one sample with the way you proposed. I am supposed to generate N samples like z from the original data y.
> - - - - - - - - - -
> p = bsxfun(@plus,ceil(5*rand(5,1,N)),5*floor(5*rand(5,5,N)));
> z = reshape(y(p),5,5,N);
>
> This should generate a three-dimensional array z of size 5 x 5 x N with elements from y. Each of the slices z(:,:,k) from k = 1 to k = N should be a valid "sample" from y in accordance with the random selection procedure you described.
>
> (This assumes that my interpretation of what you wrote is correct. Bruno had a different interpretation of your words, so you should decide which is the one you want to use.)
>
> Roger Stafford