How to handle 'NaN' values

em try to write code to find 'Interpolation error' but in Error metrix there occuring NaN at many place, even i have converted them in 'double()' now i want NaN value would replace by '0'or by its right value, Instead of NaN. plz help me . em trying to fix it bt it doesn't work.

{double(errors);

for i=1:1:size(errors, 1) for j=1:1:size(errors, 2) if errors(i, j)== 'NaN' errors(i, j)=0; else errors(i, j)=(errors(i, j)); end} end end this code is not working... i want not to occur ant NaN value during code processing. . . how do i write a code to avoide from such unwanted values. . .

Answer 1

[Paulo Silva]

errors(isnan(errors))=0

[Jan]

And one vote for the correct answer reduced to the necesary minimum.

Answer 2

[Jan]

This cannot work:

if errors(i, j)== 'NaN'

"error(i, j)" is a scalar DOUBLE, 'NaN' is a [1 x 3] CHAR vector, also known as "string". Even this would fail:

if errors(i, j) == NaN

because IEEE conventions state, that a comparison to NaN replies FALSE ever:

1 < NaN     % >> 0
1 == NaN    % >> 0
1 > NaN     % >> 0
NaN == NaN  % >> 0

Therefore isnan() must be used as shown in Paulo's answer.

[Talat]

oh Thanks guy. . .

[Paulo Silva]

+1 vote for such detailed answer :)