Sudden changes in data values - how to detect?

I have a matrix of data with around 2000 rows and only 1 column. They first few hundred are all numerical values between 1-10 and then all of a sudden, there is a change in the values to numbers between 1000-10000. How could i write a code to find the position in which this change occurs? I have been stuck with the question for several days and i can't think of a logical way to do it. I am extremely new to matlab.
Thank you

 Accepted Answer

The easiest way:
x = [randi(10, 20, 1); randi([1000 10000], 20, 1)]; % Created Data
jump = find(diff([0; x]) > 10, 1, 'first'); % Index
The ‘jump’ variable has the first index of the difference in the values that is greater than 10. Run my simulated code — and experiment with it to see how it works — then apply it to your vector.

4 Comments

Hello, thank you for your reply, I understand how this works but is there a way to standardise this procedure. For example, if I have some other data that is like in 0.1 and then all of a sudden becomes 0.00001? Is there a way I could write a code which would find a sudden change using any data?
Thank you again
My pleasure.
You would have to adjust the threshold and the comparison condition, but the code should work otherwise.
The same code for your new vector might be:
x = [0.1*randi(10, 20, 1); 1E-5*randi(10, 20, 1)]; % Created Data
jump = find(x < 1E-2, 1, 'first'); % Index
Experiment to get the result you need.
This could help me, but is it possible to show the location of multiple changes? Say you had several large jumps in one column of data...is it possible to have matlab tell us more than just the first location? Thanks!
Yes.
See my Comment to dpb’s Answer in your original Question.

Sign in to comment.

More Answers (1)

Since this question was asked the ischange function has been added to MATLAB. Its short description is "Find abrupt changes in data".

Categories

Find more on Loops and Conditional Statements in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!