| T O P I C R E V I E W |
| jbc1055 |
Posted - 01/29/2011 : 8:34:49 PM
Origin Ver. and Service Release (Select Help-->About Origin):
Operating System:
Hi,
I have faced a same problem which had been asked before: http://www.originlab.com/forum/topic.asp?TOPIC_ID=6467
I tried the function fitting organizer to define the function, but I am not sure how to give the Parameter Initialization. By the way, the equation of Varshni's law is
ET = E0 - A*Tmp^2/(Tmp+B)
Both ET and Temp are variables, and the unknown parameters that I want to approach are E0, A, and B.
Can anyone help me? Thanks a lot!
|
| 9 L A T E S T R E P L I E S (Newest First) |
| cc261 |
Posted - 10/27/2012 : 09:46:12 AM
What's matter with your 1stOpt? I never meet the situation you met. |
| grace_cmt |
Posted - 10/26/2012 : 10:50:26 PM
Hi cc261, I've taken your advice on trying to fit Varshni model using 1stopt. However, I am having problem to start with this software as the command window doesn't come out. Could you please share with me on how to do it?
Thank you!
Grace |
| cc261 |
Posted - 02/09/2011 : 04:51:18 AM
You may try 1stOpt or Auto2Fit, no need to guess initial values and the result is unique:
Variable Tmp,ET;
Function ET = E0 - A*Tmp^2/(Tmp+B);
Data;
Tmp ET
10 2.80389
20 2.8038
30 2.80341
40 2.80248
50 2.80042
60 2.7984
70 2.79599
80 2.79298
90 2.78954
100 2.78627
120 2.77785
140 2.768
Outcomes:
Root of Mean Square Error (RMSE): 0.000312632982910953
Residual Sum of Squares (SSE): 1.1728725840456E-6
Correlation Coef. (R): 0.999591049413237
R-Square: 0.999182266067056
Parameters Best Estimate
---------- -------------
e0 2.80473088708022
a -0.00348038456228686
b -1990.60746980054
 |
| larry_lan |
Posted - 01/31/2011 : 01:45:03 AM
Hi:
Now that it's not a parameter initialization issue, then you don't need to simulate it anymore. Please read this tutorial for how to create a user-defined fitting function.
Thanks
Larry
OriginLab |
| jbc1055 |
Posted - 01/30/2011 : 11:50:47 PM
Hi Larry,
Thanks a lot. The plot seems much better.
Unfortunately, I have no idea how you do that......
This is how I tried,
Step 1. Enter the function in the Fitting Function Organizer. I put E0, A, and B as parameters, and eliminate the double E0 = 1; code in Parameter Settings.
Step 2. Initialize the parameters as you mentioned.
Step3. Use the Simulate Curve option.
Step 4. Here is where I confused. Do I have to put the parameters E0, A, B myself? And the x min and x max should be 10 and 140, respectively, right?
I noticed that you initialized the parameters as: E0=2.8, A=0.0001, B=0.0001, but the outcomes are: E0=2.80492, A=600.5455 (Is the unit 1e-4?), B=3.19999E?.
So I just wonder how could I use the Simulative Curve tool?
I really appreciate your kindly help!!
|
| larry_lan |
Posted - 01/30/2011 : 10:49:50 PM
Hi:
Obviously, the model is over-parameterized, so there could be many sets of results. I initialize the parameters as: E0=2.8, A=0.0001, B=0.0001, and I got the below result.
Thanks
Larry |
| jbc1055 |
Posted - 01/30/2011 : 04:02:36 AM
Hi Larry,
I just realize that are you talking about guessing the parameters(E0, A, B) values myself by fitting the function? But I am thinking if it is possible that the software could fit the given data and provide parameters?
Thanks a lot. |
| jbc1055 |
Posted - 01/30/2011 : 03:31:52 AM
Thanks for replying!
The data and graph that I want to fit are
Tmp ET
K eV
10 2.80389
20 2.8038
30 2.80341
40 2.80248
50 2.80042
60 2.7984
70 2.79599
80 2.79298
90 2.78954
100 2.78627
120 2.77785
140 2.768
And after I followed your instruction,
the graph will be fitted like
which is not what I want. The fitting curve supposes to look similar to the origin curve.....and the parameters shown above are just not right at all. Could you give me further instruction?
I really appreciate your help. |
| larry_lan |
Posted - 01/30/2011 : 02:34:34 AM
Hi:
After you created the function, you can try the Simulate Curve dialog from Analysis: Fitting: Simulate Curve menu, and then play with this function to guess the initial values.
Thanks
Larry
OriginLab |
