Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.
|
|
|
|
Re: Question about bootstrapping to increase sample size
Posted:
Feb 22, 2013 4:51 PM
|
|
> Hello. I have data that is 45 trials, each 60 samples long. I have heard
> bootstrapping can resample it to give "new" data that can be used in
> classifier training. Can I just double-check that this is the correct way
> to do it:
>
> bootx=bootstrp(500, 'mean', x);
>
> where x is my original data and I want 500 new rows generated through
> bootstrapping. Thanks.
This doesn't really increase the sample size. If you randomly resample from
your data and append it to the data as if it were new, that would
artificially increase the sample size and make your results look more
accurate then they are.
Your function randomly applies the "mean" function to 500 new samples taken
from x. It gives you an idea of how uncertain your estimate of the mean of x
is. The result bootx will not look like x, but instead have a smaller
standard deviation centered around the mean of x.
Sadly I don't have any experience with using bootstrapping in connection
with classifier training, so I can't offer anything special there.
-- Tom
|
|
|
|
