Resampling with replacement
in [Matlab]
|
From: Peter Perkins on 1 Dec 2009 11:06 Mohammed Khalilia wrote: > Since randsample takes only vectors for the data and the weight, I had to reshape a > a1 = reshape(a,1,8); > w1 = reshape(w,1,8); > y=randsample(a1,8,true,w1); This is not "produc[ing] a bootstrap replicate of the original data according to the probability W for every instance". You're giving each separate _value_ a weight, and resampling them individually. I would think you would want to resample rows, i.e., "instances".
From: Mohammed Khalilia on 1 Dec 2009 13:00 Peter Perkins <Peter.Perkins(a)MathRemoveThisWorks.com> wrote in message <hf3eut$jt2$1(a)fred.mathworks.com>... > Mohammed Khalilia wrote: > > > Since randsample takes only vectors for the data and the weight, I had to reshape a > > a1 = reshape(a,1,8); > > w1 = reshape(w,1,8); > > y=randsample(a1,8,true,w1); > > This is not "produc[ing] a bootstrap replicate of the original data according to the probability W for every instance". You're giving each separate _value_ a weight, and resampling them individually. I would think you would want to resample rows, i.e., "instances". You are correct, I need to resample rows (instances). Here is the exact problem: "Produce a bootstrap replicate of the original data set according to the probability W for every instance i by resampling with replacement from the original data set." Initial probability W = 1/N where N is number of instances, W is 1xN vector. I think the solution provided by Lorenzo Guerrasio make sense (Thanks Lorenzo). a = [1 2;5 6;7 8;34 5]; w = [0.25 0.25 0.25 0.25]; % This was wrong above, it was 4x2 matrix ind=1:size(a,1); n_sample=4; Rind=randsample(ind,n_sample,true,w); new_sample=a(Rind,:);
From: Lorenzo Guerrasio on 1 Dec 2009 13:15 note that if w=1/N and n_sample is equal to the element of a, than the easiest thing to do is to use bootstrp [mist, Sindex]=bootstrp(1,[],a); newsample=a(Sindex,:); "Mohammed Khalilia" <mohammedsk(a)gmail.com> wrote in message <hf3ljo$g5t$1(a)fred.mathworks.com>... > Peter Perkins <Peter.Perkins(a)MathRemoveThisWorks.com> wrote in message <hf3eut$jt2$1(a)fred.mathworks.com>... > > Mohammed Khalilia wrote: > > > > > Since randsample takes only vectors for the data and the weight, I had to reshape a > > > a1 = reshape(a,1,8); > > > w1 = reshape(w,1,8); > > > y=randsample(a1,8,true,w1); > > > > This is not "produc[ing] a bootstrap replicate of the original data according to the probability W for every instance". You're giving each separate _value_ a weight, and resampling them individually. I would think you would want to resample rows, i.e., "instances". > > You are correct, I need to resample rows (instances). Here is the exact problem: > "Produce a bootstrap replicate of the original data set according to the probability W for > every instance i by resampling with replacement from the original data set." > > Initial probability W = 1/N where N is number of instances, W is 1xN vector. > > I think the solution provided by Lorenzo Guerrasio make sense (Thanks Lorenzo). > > a = [1 2;5 6;7 8;34 5]; > w = [0.25 0.25 0.25 0.25]; % This was wrong above, it was 4x2 matrix > ind=1:size(a,1); > n_sample=4; > Rind=randsample(ind,n_sample,true,w); > new_sample=a(Rind,:);
From: Mohammed Khalilia on 1 Dec 2009 13:28 "Lorenzo Guerrasio" <lorenzo.guerrasio(a)email.it> wrote in message <hf3mfu$bl1$1(a)fred.mathworks.com>... > note that if w=1/N and n_sample is equal to the element of a, than the easiest thing to do is to use bootstrp > [mist, Sindex]=bootstrp(1,[],a); > newsample=a(Sindex,:); > > > > > "Mohammed Khalilia" <mohammedsk(a)gmail.com> wrote in message <hf3ljo$g5t$1(a)fred.mathworks.com>... > > Peter Perkins <Peter.Perkins(a)MathRemoveThisWorks.com> wrote in message <hf3eut$jt2$1(a)fred.mathworks.com>... > > > Mohammed Khalilia wrote: > > > > > > > Since randsample takes only vectors for the data and the weight, I had to reshape a > > > > a1 = reshape(a,1,8); > > > > w1 = reshape(w,1,8); > > > > y=randsample(a1,8,true,w1); > > > > > > This is not "produc[ing] a bootstrap replicate of the original data according to the probability W for every instance". You're giving each separate _value_ a weight, and resampling them individually. I would think you would want to resample rows, i.e., "instances". > > > > You are correct, I need to resample rows (instances). Here is the exact problem: > > "Produce a bootstrap replicate of the original data set according to the probability W for > > every instance i by resampling with replacement from the original data set." > > > > Initial probability W = 1/N where N is number of instances, W is 1xN vector. > > > > I think the solution provided by Lorenzo Guerrasio make sense (Thanks Lorenzo). > > > > a = [1 2;5 6;7 8;34 5]; > > w = [0.25 0.25 0.25 0.25]; % This was wrong above, it was 4x2 matrix > > ind=1:size(a,1); > > n_sample=4; > > Rind=randsample(ind,n_sample,true,w); > > new_sample=a(Rind,:); W = 1/N is only the initial value, then W will change in every iteration.
First
|
Prev
|
Pages: 1 2 Prev: mincx to gevp Next: Converting Binary images/Grayscale images to move frames |