| Author |
 Topic  |
|
|
jcarifin
Taiwan
6 Posts |
 Posted - 03/29/2023 : 2:34:30 PM
|
Origin Ver. and Service Release (Select Help-->About Origin): 2016 (64 bits)
Operating System: Windows
Hello, I have a question about the meaning of standard error in curve fitting.
I fitted some histograms with GaussAmp function. And the Origin gives me the result including area (the area under the function?). It also gave me standard error (SE).
What does this error mean? So, suppose I want to compare the area of several fittings from several histograms:
Function 1 (Area +- SE): 190 +- 12
Function 2 (Area +- SE): 238 +- 21
Function 3 (Area +- SE): 200 +- 13
Can I treat them as mean +- SD, and do some statistical tests similar to T-test and ANOVA to find out the P value or significanct difference? If yes, how to do it in Origin?
Thank you in advanced!
|
Edited by - jcarifin on 03/29/2023 2:39:54 PM |
|
|
YimingChen
1651 Posts |
|
|
jcarifin
Taiwan
6 Posts |
Posted - 03/29/2023 : 5:24:06 PM
|
Dear James,
Thank you for the quick reply, but from my understanding, what I want to do is different with the app.
What I wanted to compare are not datasets, but several values with known SE obtained from Origin's curve fitting: how significantly different they are and find the P-value (like a T-test or ANOVA). If it were datasets, Origin can do it easily.
Another problem was, I am not sure about the meaning of Standard Error obtained from the area calculation of Gaussian curve fitting in Origin; whether I can treat it like a Standard Devation (SD), and use it to find the P-value from 2 means with known SD or SEM. Like, if we can treat it as SD or SEM, what is the N number.
From my understanding and after trying the app, I think the app is used to compare datasets to some model/function, which is similar to curve fitting.
Please correct me if I am wrong. |
Edited by - jcarifin on 03/29/2023 7:32:19 PM |
 |
|
|
Sam Fang
293 Posts |
Posted - 03/30/2023 : 11:40:32 PM
|
Hi jcarifin,
We suggested you use Compare Datasets and Fit Parameters app. It is the general method. You should choose Gauss or Gaussian function as this app does not support to compare derived parameters. In the app you can only choose area parameter to compare (assume other parameters are same in two data), and use F-test. F-test will compare residuals in two models (area parameters are same for two data, and different for two data).
Standard error in Origin's curve fitting is the standard deviation of best fit parameters if you repeated the fit many times, e.g. use bootstrap.
If you insisted on using t-Test though it is not recommended, you can choose Origin's Two-Sample t-Test, choose Summarized Data as Input. You can provide fake SD and sample size. Suppose df and se is degrees of freedom and standard error of the parameter in fitting, fake sample size is df+1, fake SD=se*sqrt(df+1)
Thanks.
Sam
Sam
OriginLab Technical Services |
Edited by - Sam Fang on 03/31/2023 01:55:50 AM |
|
|
|
jcarifin
Taiwan
6 Posts |
Posted - 03/31/2023 : 3:57:34 PM
|
Dear Sam,
Thank you very much for the reply! Now I have a clear understanding of the SE and what to do with the function.
Kind regards,
Jcarifin |
|
|
| |
Topic |
|
|
|
Standard error in curve fitting