Matlab delete's value's from array

Hi,

I have a code in matlab that generates an big array [3648x1]. This array looks something like:

[NaN
NaN
NaN
NaN
0 
0 
1 
0
1 
1 
NaN
NaN
NaN
0 
0 
1  
1 ]

I want to replace the zeros with NaN, where after I can delete all the NaN's

This is the code, where variable1 is the [3648x1] array (and a, b, c are also [3648x1] array's);

for a = 0 : (length(variable1)-1);
    if variable1(a+1) == 0;
        variabele2(a+1) = NaN;
    end
end

k = [variabele2, a, b, c];

k(any(isnan(k),2),:)=[];

variabele2 = k(1:end,1);
a = k(1:end,2);
b = k(1:end,3);
c = k(1:end,4);

But after the for loop, the array is reduced to an [1x1656] array. How is this possible? And more important, how can I fix it so the code will do what I want.

[loes visser]

The data is time related, thats why I want to delete a whole row when a value in variable1 is NaN or 0. Then I can make new variables were variable1(1) has still the same timestamp as a(1) or b(1).

This code is working with another variable, I dont get why it wouldnt work with this variable and why it is reducing its values.

Answer 1

[Mischa Kim]

loes, based on your description, how about

 mat(isnan(mat) | mat==0) = [];

where mat is, for example, the matrix you show at the very top.

[loes visser]

I thought this might work, but there might be zeros in a, b, and c as well. I only want to delete the rows where variable1 is NaN or 0. For example, matrix k = [variabele2, a, b, c] might look like;

1     20   3   0
NaN   19   2   1.1
NaN   20   1   0.9
NaN   19   2   0.4
0     18   0   0.3
1     19   1   0.8
1     20   3   1.0
0     21   1   1.1
1     22   0   0.7
NaN   18   1   0.5
NaN   19   3   0.9

So;

 k(isnan(k) | k==0) = [];

Will delete row 1 and 9 as well The data is time related, thats why I want to delete a whole row when a value in variable1 is NaN or 0. Then I can make new variables were variable1(1) has still the same timestamp as a(1) or b(1).

[loes visser]

It worked with a double expression.

variabele1(any(variabele1== 0,2),:)= NaN;  

k = [variabele1, a, b, c];

k(any(isnan(k),2),:)=[];

variabele2 = k(1:end,1);
a = k(1:end,2);
b = k(1:end,3);
c = k(1:end,4);

Thank you!

[Mischa Kim]

Correct. If k is 2D use

k([isnan(k(:,1)) | k(:,1)==0],:) = []

[Stephen23]

the square brackets are not required:

 k(isnan(k(:,1)) | k(:,1)==0,:) = []

[Mischa Kim]

I like to use them at times to help with readability.

Answer 2

[Matthew Eicholtz]

Suppose you have an M-by-N matrix (A) of NaNs, 0s, and 1s:

M = 3648; % number of rows
N = 4; % number of columns

A = double(rand(M,N)>0.1); % random mix of 0s and 1s
A(rand(size(A))>0.9) = NaN; % randomly add some NaNs

You can create a mask matching the conditions you are looking for (in this case, we only want to keep rows that do not have NaNs or 0s):

mask = all(~isnan(A) & A~=0,2);

Then, grab the rows corresponding to the mask.

B = A(mask,:);

Answer 3

[Thorsten]

You remove all rows that contain a NaN in the line

k(any(isnan(k),2),:)=[];

and then you assign the variables

 variabele2 = k(1:end,1);
 a = k(1:end,2);
 b = k(1:end,3);
 c = k(1:end,4);

so afterwards the variables are smaller. So what's the problem?

Answer 4

[Andrei Bobrov]

k = [1     20   3   0
NaN   19   2   1.1
NaN   20   1   0.9
NaN   19   2   0.4
0     18   0   0.3
1     19   1   0.8
1     20   3   1.0
0     21   1   1.1
1     22   0   0.7
NaN   18   1   0.5
NaN   19   3   0.9];

out = k(k(:,1)~=0 & ~isnan(k(:,1)),:);