| Author |
 Topic  |
|
|
supermarche
2 Posts |
 Posted - 12/06/2008 : 8:14:48 PM
|
Origin Ver. and SR (Select Help-->About Origin): Origin 8 SR4
Operating System: Windows XP Pro SP3
Hi,
I've got over 20 similar data sets. They each look like an approximation of the heaviside / theta function. I want to make a linear fit for both plateaus and the ascending line. Afterwards it is necessary to determine the middle of the ascending line. How can I fulfill this task more effecient than dividing each dataset into 3 parts (plateau, ascending, plateau) and doing 3 independend linear fits?
thanks
Marc |
|
|
Deanna
China
Posts |
Posted - 12/08/2008 : 02:34:42 AM
|
Hi Marc,
The Linear Fit tool of Origin 8 is capable of performing fitting to multiple ranges at the same time. But, I am not sure how you separate the plateus and the ascending line. If you know the range of each plateus and the ascending line, I think we can utilize the Linear Fit tool and maybe, the Analysis Template feature to automate the fitting.
Deanna
OriginLab Technical Services |
 |
|
|
supermarche
2 Posts |
Posted - 12/08/2008 : 3:08:40 PM
|
Hi,
I know the Ordinates of the first and second Plateau:
first: y < -2*10^-4
second: y > 10^-3
I only found a limitation by x-values for linear fits, but this is not applicable, because the data varies for the x-values. It would be nice, if you could give me further tips.
Bye,
Marc |
|
|
|
liuxiaopi
China
Posts |
Posted - 12/09/2008 : 05:17:07 AM
|
Hi, Marc,
You can try to define a three segment piece-linear function as follows:
y= (y1*(T1-t)+y2*(t-t1))/(T1-t1) t1<t<T1
y= (y2*(T2-t)+y3*(t-T1))/(T2-T1) T1<t<T2
y= (y3*(t3-t)+y4*(t-T2))/(t3-T1) T2<t<t3
Note 1:
t indep;
y dep;
Fitted parameters: T1,T2, the start and end t of the slope
y1 , y2, y3, y4;
y1 is the start y value;
y2 is the start of the slope y value
y3 is the end of the slope y value
y4 is the end y value;
Note 2:
t1 and t3 are are min and max of your x data range, and you can use initial code in FDF (Originlab 8.0) to initilize them. And Set Fixed status of these two parameters in the fitting to avoid over-parametrized.
Note 3:
The convenience of this math formula is that: you can get contact y value for these three region (plateaus and the slope).
For example, (y2+y3)/2 can give you the middle value you are intesested.
hope this helps,
Thanks,
Jack
|
|
|
| |
Topic |
|
|
|
possible multiline fitting ?