Best way to get rid of/ prevent function from creating complex numbers?

Hi,

I have a function called dfaedit which somehow sometimes gives me complex numbers as outputs. I have not found a good way to deal with this except for deleting the complex numbers all together. My online search didn't yield any useful results.

What is the best way to approach this? I would like to keep all the information in my data and prevent the function from creating complex numbers.

The code I am running is:

for k=1:length(nonglitch_xyz_h_ref)
   Hurst_nonglitch_h_ref(k)=dfaedit(nonglitch_xyz_h_ref{k},0,0,0);
end

Thanks!

[Matt J]

The question is under-formulated There are many ways you could convert complex numbers to real numbers without deleting them. You could throw away the imaginary part,

x=real(x)

You could take it's absolute value,

x=abs(x)

What does it mean to "keep all the information" - which aspect of the data do you deem the information and which do you deem expendable?

[lil brain]

Does throwing away the imaginary part change its value? I essentially want to keep them as comparable to non.imaginary numbers as possible. Thanks!

[Stephen23]

"Does throwing away the imaginary part change its value? "

Yes: its value includes both the real and imaginary parts, so when you take only the real part you are "changing" its value.

"What is the best way to approach this?"

Find out why your code returns complex values.

Answer 1

[John D'Errico]

You cannot simply force that a complex number will never arise. There is no flag you can just set to tell MATLAB to never create a complex number. For example, what if you have a square root in your computations, and the argument just happens to be negative?

Can you write code that is more robust against this happening? Well, yes. But that requires you write careful, high quality code. It will require you to know what operations can create a complex result, and how to deal with that eventuality. Just part of being a good programmer.

Answer 2

[David Goodmanson]

Hello Jeffrey,

Deleting unsatisfactory results is of course not a good way to deal with this. I ran case {15} from your matfile data. In this case log_Q has 49 elements from 49 iterations. The first two entries are complex, the rest are real. We have

log_Q = log10(sqrt(s/den))

and the culprit is s, whose value is accumulated in a for loop for each of the 49 iterations. The first two iteration values are negative, the other 47 are positive, hence the complex result. So now you get to go back and find out why the first two s values come out negative. There might be an error, or maybe you will decide that the algorithm does not really apply in the first couple of iterations for reason such-and-such and eliminate those values, which is far better than deleting them ex post facto.

Answer 3

[Akira Agata]

How about using isreal function?