What should I do for NaN values?

I have an 1x507 array with NaN values and I am using this array for another calculation. That's why I think these are effecting my calculation. What should I do about them? Should they stay as they are or I should do something?

[C.J. Harris]

Without know what calculation you are doing, only you can answer that. If the calculation is "How many NaN values are there in a 1x507 array?", then keep them, if you are doing some numeric calculation then obviously you can't use the NaN values.

But also, you can't just delete them without understanding why they are there in the first place - why does your source data have NaN values?

[Sukru Yavuz]

I need them for my calculation. I don't need to count them. I need to calculate the sums of all 507 columns and save them as a table. Which is again equal to NaN. I think the numbers are very very very lower than zero that's why they are NaN.

likelihood(nn) = multiplycolumns(nn) * bolumler{nn}; (After this step the array has multiple NaN and Inf.) likelihoodsums = sum(likelihood)

[Stephen23]

"What should I do about them? Should they stay as they are or I should do something?"

Why do you assume that you have to "do something" about them? NaN values are there for a reason, and it is usually a bad strategy to turn NaN values into some fake data values (e.g. 0): that would probably distort your calculations even more.