Finding a point in the data set after which the slope of the line is approximately zero.
shivatarun17
Mon, 02/13/2017 - 01:51 pm
Hi! I am having a real time data set and I need to find the point from which the slope is decreasing at a constant rate (i.e the data is at a very small angle with X axis) Can anyone help me to find that point. Please see the attached figure and the wave data set.
March 9, 2017 at 07:32 pm - Permalink
You could do a 'sliding' fit to a constant function ( f(x)=y0 ; like box car averaging ) in a small range of your data ( [x0, x0+window] ) and find the minimum of the fit error as a function of x0.
I'd be careful with derivatives in your case since there is significant noise in your data.
Furthermore, you ask for a 'point' where it starts. Since there is noise, how do you define that point? More precise: how do you distinguish between a data point that is in that 'flat area' by noise and one that is really there? Think about your X error bars.
The previously mentioned intersection of functions is really ONE point (if not pathological). But you might need to imply a model to justify this definition.
HJ
March 10, 2017 at 02:10 am - Permalink