find for sign changes

Columns 49 through 56

    0.8710    0.9483    1.0408    1.1538    1.2956    1.4796    1.7292    2.0885

Columns 57 through 64

    2.6538    3.6815    6.1847  119.1838 -118.8424   -5.8436   -3.3407   -2.3137

Above is part of my results. I need to find for value that changes from positive to negative. In the above data, after 119.1838(column 60) the -118.8424(column 61) appears. I need to find this value. Hoe can I do this?

Answer 1

[Walter Roberson]

find(x(1:end-1)>0 & x(2:end) < 0)

Another approach that might be close enough for your purposes:

find(diff(x>=0),1)

[Jan]

"From positive" means >=0 instead of >0.

[Walter Roberson]

Wikipedia: "In mathematics, every real number other than zero is either positive or negative"

In other words, zero is not positive so "from positive" means >0

[Jan]

But what happens with your first approach for [1, 0, -1]?

The OP must decide this detail.

[Walter Roberson]

In the case of [1, 0, -1] there is no position at which the value is positive and the next value is negative :)

Answer 2

[Jan]

Walter's method detects multiple sign changes also. This finds the first negative number only, but this meets your problem description also:

index = find(x < 0, 1, 'first')

Answer 3

[Rasmus Herlo]

Finding any place where the sign changes, either from neg to pos or from pos to neg could also be done in general by this:

Idx_Change = sort([strfind(x>=0, [0 1]) strfind(x>=0, [1 0])]);

[Walter Roberson]

Yes but be careful about exact 0.

Answer 4

[Alireza Ahani]

@Walter Roberson is right,

but when I checked with a dataset that includes zero-down-crosing event, they would be ignored, so I added some lines to his:

ZDCI = find(x(1:end-1)>0 & x(2:end) < 0); % zero down crossing index
Zero_indx = find(x == 0);
k = find( x(Zero_indx-1) > 0);
ZDCI = sort([ZDCI Zero_indx(k)]);

[Walter Roberson]

If the task is zero down crossing and zeros can be encountered, then the above code is not correct, as it does not take into account multiple zeros. For example for zero down crossing, you might have +1 followed by 5 zeros, followed by -1. The zeros could, for example, be present due to filtering of values that were close to zero, such as due to measurement noise on a sigmoid that crossed zero. Detection for such cases cannot be restricted to any fixed finite number of look-back.