add and remove columns from a marix

I am writing a code , in fact I have a matrix and I select the fist column then I add next column and in each epoch one column will be adding to the previous for example first 1 , the second epoch 1 and 2 , the third epoch 1, 2 and 3 and so on .

I do that like this

for i = 1:size(Matrix,2) new matrix after each iteration = Matrix(:,1:i);

but now before i put end, I would like to evaluate this new matrix.

I would like to say if nnn> 0.5 then keep this and continue the loop otherwise remove this column from the Matrix and go for the next one and continue the loop

Do you have any idea?

Answer 1

[Fangjun Jiang]

n=rand(5,1);
Matrix=magic(5);
NewMatrix=[];
for k=1:size(Matrix,2)
    if n(k)>0.5
    NewMatrix(:,end+1)=Matrix(:,k);
    end
end

[Niki]

In fact, i want to make a regression each time

then i see if the add column can improve i keep it otherwise i remove it.

I would like to say if RMSE< 0.5 then keep this and continue the loop otherwise remove this column from the Matrix and go for the next one and continue the loop

do you think your new like will do that ?

[Niki]

i think it is a good example

[XL,YL,XS,YS,BETA,PCTVAR,MSE,STATS] = PLSREGRESS(X,Y,1)

if MSE <0.5 keep the column and continue the loop

if the MSE >0.5 remove the column from the matrix and continue with the next column

[Fangjun Jiang]

If you want to remove the column, do the follow. Notice the loop from 5 to 1. This is because if you remove the column, the size of the matrix is changed.

n=rand(5,1);

Matrix=magic(5);

for k=size(Matrix,2):-1:1

if n(k)>0.5

Matrix(:,k)=[];

end

end

[Niki]

I would like to have only one column first and then add the next one not have all and then remove one by one in each iteration ,

do you know what I mean?

[Fangjun Jiang]

Then it's the code in the answer.

[Niki]

in fact I did like this

X=rand(10);

y=rand(10,1);

[n,p]=size(X);

RMSECF=[];

for i=p:-1:1;

[xl,yl,xs,ys,beta,pctvar,mse] = plsregress(X(:,1:i),y,1);

if MSE<0.5

X(:,i)=[];

end

end

[Fangjun Jiang]

I can't comment on your solution, because I don't have your plsregress() function. RMSECF is not used. Not sure if mse and MSE mean the same.

[Niki]

the plsregress is a Matlab function, which version do you have?

please just doc it , I revised it

but the mse seems strage, it should be 10 numbers and there is only 4 which I can not recognize them

X=rand(10);

y=rand(10,1);

[n,p]=size(X);

mse=[];

for i=p:-1:1;

[xl,yl,xs,ys,beta,pctvar,mse] = plsregress(X(:,1:i),y,1);

if mse<0.5

X(:,i)=[];

end

end

[Fangjun Jiang]

The function is in the Statistics Toolbox which I don't have.

[Niki]

You can use of following function , please

http://www.mathworks.com/matlabcentral/fileexchange/16465-pls/content/PLSI.m

[Fangjun Jiang]

Sorry, I won't do it. It seems to be a different topic. Ask a different question and someone with statistics expertise might be able to help you.

[Niki]

Thanks Fangiun