Find max/min while ignoring data spikes

My problem is that I want to ignore the sudden dropoff in the attached plot. I'm using the following to get the minimum and maximum:

ErrPriMax = ErrorPri<1;

ErrorPriMax = max(ErrorPri(ErrPriMax));

ErrPriMin = ErrorPri>-.4;

ErrorPriMin = min(ErrorPri(ErrPriMin));

ErrRedMax = ErrorRed<1;

ErrorRedMax = max(ErrorRed(ErrRedMax));

ErrRedMin = ErrorRed>-.4;

ErrorRedMin = min(ErrorRed(ErrRedMin));

The correct minimums should be around -.2 for the primary error and -.05 for the redundant error; however, I am getting around -.4 for both since that is the hard limit I coded in. Is there a better way to do this? I essentially want to ignore the two spikes.

ErrorPri and ErrorRed are the datasets I am working with along with Time. They dont go to infinite, they go more closely to -20 (which is far greater than reality).

Any help is greatly appreciated.

Thank You!

Answer 1

[JohnGalt]

it looks like '-1' is some specific value which you want to ignore... so you remove all '-1's from the array before you calculate the min e.g.

   testArray = [.1 .2. -.1 .2 .1 .2 -1 .2. .1];
   testArray(testArray<=-1)==[];
   disp(testArray) % which will be: [.1 .2. .1 .2 .1 .2 .2. .1];
   disp(min(testArray))  % which will be -.1

Just be aware that you have removed a point from your original array so you have changed the length of the original

[Marcus Solis]

The only problem with that is that the data goes to -1. So the array looks more like [-0.87 -0.91 -0.95 -0.97 -0.99 -1.02...]. So when I add a filter like the one you have above, the min is going to be -0.99. I was thinking that I could take the derivative, and if it goes close to infinity then I know there is a spike. But even doing that I would have to compare the derivative across 30 to 50 points.