How can one refine a trace that has separate trendlines for several sections? As an example, I generated two noisy lines as pasted below, x=1:20 for section 1, x=21:50 for section 2, and smoothed their joint part. This is an example and the trendline may not be linear. Is there a way to do advanced fit of the whole pattern to get the right trendline? Thanks a lot. x y 1 0.55002 2 0.69863 3 0.41885 4 0.73308 5 0.74913 6 0.95625 7 1.04946 8 0.95814 9 1.11419 10 1.16559 11 1.48408 12 1.33935 13 1.41326 14 1.83436 15 1.85974 16 1.87008 17 1.77542 18 1.97046 19 1.94242 20 1.90431 21 1.9532 22 1.8583 23 1.68719 24 1.5434 25 1.4449 26 1.47086 27 1.28461 28 1.16005 29 0.90488 30 0.80683 31 -0.07985 32 -0.1817 33 -0.43662 34 0.00805 35 -0.94654 36 -0.97525 37 -0.95891 38 -1.46769 39 -1.19232 40 -1.74449 41 -1.35926142 42 -2.295079242 43 -1.961904251 44 -1.850321692 45 -2.998005976 46 -2.405363235 47 -3.061000348 48 -3.073086682 49 -3.772977692 50 -3.402583242
Maybe it is better to fit two sections separately. You can use the range selector to select different sections and do the fitting with different functions.
I'm going to dust off a bit of ancient LabTalk wisdom for this one. The LR object has an option to search the beginning or the end of a dataset for a linear pattern given a criteria controlled by MAXLRANGE and CHKLINEARR. Consult the Help for more information.
Here is a script that uses the default values to determine the linearity of the beginning and ending of your data:
lr -b %C; type Slope in the beginning \x3D $(LR.B), Intercept \x3D $(LR.A); lr -e %C; type Slope at the end \x3D $(LR.B), Intercept \x3D $(LR.A);
which outputs:
Slope in the beginning = 0.09392, Intercept = 0.33885 Slope at the end = -0.18633, Intercept = 5.82756
Section fitting?
Posted - 11/06/2007 : 12:24:25 PM
