Using sign(x) with -0.00?

Hi, I need to go through a vector and check the sign of an oscillatory function but I have a lot of -0.0000's at the start which is going to mess up what I need to do.

Is there anyway to convert them to 0.000 rather than -0.000?

Thanks

Answer 1

[Jan]

Matlab does not have a -0.0, therefore what you are seeing must be a -0.000000000000001 or similar, which is hidden by the display format. To see more digits type this in the command window:

format long g

If you have good reasons to claim, that these data are pure noise, you can smooth it:

x(abs(x) < 1e-10) = 0.0;

But be sure to note the physical motivation of the threshold value.

[Martin]

cheers, this worked a treat. I was already using the format long command. What does the extra 'g' do?

Turns out it was something like 1e-29 so pretty small!

[José-Luis]

Please accept an answer if it helps you.

[Walter Roberson]

"g" means to choose the best representation automatically.

[Jan]

Or to explain "best": the "e" notation is used, when it can display more significant digits than the 0.0000xyz notation.

[James Tursa]

To clarify what Jan wrote, IEEE double and single floating point (which MATLAB uses) do in fact have distinct -0.0 bit representations (sign bit set to 1 and all other bits reset to 0) that is different from 0.0 (all bits reset to 0). But by default MATLAB will not display the minus sign '-' for these numbers, so when you see something like -0.0000 printed to the screen it is not really a -0.0 value. Rather it represents a small negative number compared to the other numbers that were printed to the screen at the same time (there simply weren't enough digits printed to show the non-zero digits of the small negative number). To see this type the following at the MATLAB prompt:

format hex
0
-0
0 == -0
typecast(0,'uint64') == typecast(-0,'uint64')

Answer 2

[José-Luis]

your_mat(abs(your_mat) < 100*eps) = 0; %or some other limit

[Martin]

This looks like a good thing to do also, thanks. For all intents and purposes the numbers I am getting -0.00000 for could be equated to zero.