Just choose some point where the curve should have settled into its long term nominal state. That should be at or about a time of 25 in your plot.
Next, take the median value of the curve over that interval, here from a time of 25 to the end.
help median
MEDIAN Median value.
For vectors, MEDIAN(x) is the median value of the elements in x.
For matrices, MEDIAN(X) is a row vector containing the median value
of each column. For N-D arrays, MEDIAN(X) is the median value of the
elements along the first non-singleton dimension of X.
MEDIAN(X,"all") is the median of all elements of X.
MEDIAN(X,DIM) takes the median along the dimension DIM of X.
MEDIAN(X,VECDIM) operates on the dimensions specified in the vector
VECDIM. For example, MEDIAN(X,[1 2]) operates on the elements contained
in the first and second dimensions of X.
MEDIAN(...,NANFLAG) specifies how NaN values are treated:
"includemissing" / "includenan" -
(default) The median of a vector containing NaN values
is also NaN.
"omitmissing" / "omitnan" -
The median of a vector containing NaN values is the
median of all its non-NaN elements. If all elements
are NaN, the result is NaN.
Example:
X = [1 2 4 4; 3 4 6 6; 5 6 8 8; 5 6 8 8]
median(X,1)
median(X,2)
Class support for input X:
float: double, single
integer: uint8, int8, uint16, int16, uint32, int32, uint64, int64
See also MEAN, STD, MIN, MAX, VAR, COV, MODE.
Documentation for median
doc median
Other uses of median
categorical/median prob.NormalDistribution/median
codistributed/median tabular/median
datetime/median tall/median
duration/median timeseries/median
fixedpoint/median
Finally, discard any value ABOVE that median. You can just use min to perform that task. You might even be more aggressive, and choose the 25th percentile of the curve, as perhaps a better estimate of the long term behavior, better insuring you don't get any of those outlier segments at all.
