Subtracting different matrices with NaN values

Hello everybody,

i have the following problem. There are 4 matrices (each containing 234x7690), i will call them A B C D. Then I have to calculate the result of A - B - C - D. All of the matrices have multiple Nan's. So if one of the 4 matrices has no value the result is NaN for this cell. But i need a result whenever A is a real number and at least one of the other has also a real value. So for example if (36 - Nan - 5 - NaN) the result should be 31. I hope you got my problem. Thanks for your answers.

Answer 1

[Adam]

Just replace NaNs with 0s

e.g.

B( isnan(B) ) = 0;

[Sascha Winter]

Yes, i also thought about this idea. But one of the restriction is that it should give me only a result if at least one is not NaN. So when i replace all Nan's with 0, then it may happen that B, C and D were originally NaN and it gives me a result, although i don't want a value for such a case.

[Adam]

Well, you could do some indexing to find these situations if you stack your matrices as

myMatrices = cat( 3, B, C, D );
nanLocations = isnan( myMatrices );
allNaNs = sum( nanLocations, 3 ) == 3;

then use allNaNs as a mask to your result array to set any values where it is true to NaN in your result matrix.

I don't have time to give it as a complete solution, but you should get the idea.