Set specific values to NaN

Hello everyone

My problem is as follows. I have a matrix like this for example:

A = [1 2 3 4 5 6 6 6 6; 
     2 4 3 5 4 6 7 7 7; 
     5 4 4 4 3 7 8 8 8; 
     2 1 3 3 3 3 3 3 5]'

Now i want to get a matrix which deletes the values which are similar (starting with the second) untill the end of the column. So columns 1, 2, 3 should delete some values while column 4 shouldn't delete any numbers.

At the end the matrix should look like this

A = [1 2 3 4 5 6 NaN NaN NaN; 
     2 4 3 5 4 6 7 NaN NaN; 
     5 4 4 4 3 7 8 NaN NaN; 
     2 1 3 3 3 3 3 3 5]'

I hope you get my problem. Thanks in Advance.

[Peter Strassmann]

Be aware that NaN values are only defined for data with single or double precision! It is resets integers to zero, e.g. uint8, int8.

rng(1);
A=randi([0,8], [100,1], "int8");
%A=double(A);
A(A>2)=nan;

If you want to associate NaN values to such data, you have to convert them first (before the remapping) with single(A) or double(A) into values with the according precision.

Answer 1

[Stephen23]

>> A = [1,2,5,2;2,4,4,1;3,3,4,3;4,5,4,3;5,4,3,3;6,6,7,3;6,7,8,3;6,7,8,3;6,7,8,5];
>> idx = cumprod(double(diff(flipud(A))==0),1);
>> idx(end+1,:) = false;
>> idx = logical(flipud(idx));
>> A(idx) = NaN
A =
     1     2     5     2
     2     4     4     1
     3     3     4     3
     4     5     4     3
     5     4     3     3
     6     6     7     3
   NaN     7     8     3
   NaN   NaN   NaN     3
   NaN   NaN   NaN     5

[Sascha Winter]

Thanks for your answer. It works for this matrix but it doesn't work for columns where all positions are NaN, because i get the error code: Error using logical NaN's cannot be converted to logicals. Do you have an idea how to pass this problem?

[Stephen23]

@Sascha Winter: how would you get a complete column of NaNs? As far as I can tell, if all elements in a column have the same value then following the description in your question "...starting with the second..." the first element would not be NaN. So how do you get a column of only NaNs?

>> A = [1,2,5,5;2,4,4,5;3,3,4,5;4,5,4,5;5,4,3,5;6,6,7,5;6,7,8,5;6,7,8,5;6,7,8,5]
A =
     1     2     5     5
     2     4     4     5
     3     3     4     5
     4     5     4     5
     5     4     3     5
     6     6     7     5
     6     7     8     5
     6     7     8     5
     6     7     8     5
>> idx = cumprod(double(diff(flipud(A))==0),1);
>> idx(end+1,:) = false;
>> idx = logical(flipud(idx));
>> A(idx) = NaN
A =
     1     2     5     5
     2     4     4   NaN
     3     3     4   NaN
     4     5     4   NaN
     5     4     3   NaN
     6     6     7   NaN
   NaN     7     8   NaN
   NaN   NaN   NaN   NaN
   NaN   NaN   NaN   NaN

[Sascha Winter]

You understood my the wrong way. In my original matrix (318x7690) i have columns where all values are NaN. If you take the cumprod of columns where all values are NaN it isn't possible. The columns in my original matrix contain a lot of NaN's so the cumprod function will not work. I hope you get what i mean. I give you an example of how my matrix probably looks like.

A = [123456666; 344456621; NaNNaN2456666; 122222NaNNaNNaN; NaNNaNNaNNaNNaNNaNNaNNaNNaN]'

My matrix then should look like this

A = [123456NaNNaNNaN; 344456621; NaNNaN2456NaNNaNNaN; 12NaNNaNNaNNaNNaNNaNNaN; NaNNaNNaNNaNNaNNaNNaNNaNNaN]'

So rows 3.4 and 6 wouldn't work with this solution. Thanks im Advance for your help.

[Stephen23]

Try this:

   A = [1,2,3,4,5,6,6,6,6; 3,4,4,4,5,6,6,2,1; NaN,NaN,2,4,5,6,6,6,6; 1,2,2,2,2,2,NaN,NaN,NaN; NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]'
  idx = diff(flipud(A));
  idx = cumprod(double(idx==0|isnan(idx)),1);
  idx(end+1,:) = isnan(A(1,:));
  idx = logical(flipud(idx));
  A(idx) = NaN

which outputs this:

A =
   1     3   NaN     1   NaN
   2     4   NaN     2   NaN
   3     4     2   NaN   NaN
   4     4     4   NaN   NaN
   5     5     5   NaN   NaN
   6     6     6   NaN   NaN
 NaN     6   NaN   NaN   NaN
 NaN     2   NaN   NaN   NaN
 NaN     1   NaN   NaN   NaN

[Sascha Winter]

I think it works thanks a lot.