| Author |
 Topic  |
|
|
holyone2
United Kingdom
3 Posts |
 Posted - 08/24/2017 : 11:57:03 PM
|
Origin Ver. and Service Release (Select Help-->About Origin):
9.1.0 (64-bit)
Operating System: Win 10 Pro
I performed an non-linear fit to an expression of the form:
y=exp(ln(10)k/x)-1
Where:
k is parameter (VT)
y (sens_perc_unbounded)
x (SS)
Image of result here:
https://www.dropbox.com/s/kbulky02h1hy8i4/Untitled.png?dl=0
Despite apparently a good fit the ANOVA gives a Prob>F value of 1, which I believe means far from statistically significant. As I understand it based on this: http://www.originlab.com/doc/Origin-Help/NLFit-Theory#ANOVA_Table, the value is the p-value, so as its not small I cannot reject the null hypothesis that y=constant (where constant is the mean of the y values, I read on a different website). Clearly y=constant would be a much poorer fit to this function, so I don't understand the result.
- I'd also be interested if anyone knows what the null hypothesis is for the test found at: "fitting> non-linear curve fitting> advanced > quantities to computer >t-value" ?
I'd be grateful for any help. |
Edited by - holyone2 on 08/27/2017 12:12:55 AM |
|
|
Echo_Chu
China
Posts |
Posted - 08/25/2017 : 04:11:33 AM
|
Hi,
From the F Value in your ANOVA table, the Prob value should not be 1, I am wondering whether there is a bug in the the old Origin. I can not even reproduce it in my Origin 9.1 SR2. Could you show me your About Origin dialog to let know which build of Origin you are working with. You can open the dialog from menu "Help: About Origin"
And could you try our latest version to see whether there is still problem?
http://www.originlab.com/index.aspx?go=Downloads/OriginEvaluation
In regard the t values, you can look at the explanation from the page below. In addition, in our latest version, Origin 2017, there is footnote to explain the meaning in result table.
http://www.originlab.com/doc/Origin-Help/Interpret-Regression-Result#t-Value
Echo
OriginLab Technical Service |
 |
|
|
holyone2
United Kingdom
3 Posts |
Posted - 08/27/2017 : 12:11:42 AM
|
Thank you, the documentation here is very useful. So in this example, given the equation I’m using the t-test is taking the parameter=0, so y=exp(0)-1=0 is the null hypothesis.
The documentation states the F-value null hypothesis is y=constant, but please could you confirm what that constant is? Is it the mean of the y value data? y=0.52673 for the F-test.
The build I was using was: b271 (Academic) Origin 9.1.0 (64-bit) sr2
I have now installed Origin2017 but get the same problem (b9.4.2.380 Origin 2017 (64-bit) SR2)
I’ve attached the Origin file edited in Origin 2017, worksheet:
https://www.dropbox.com/s/05zs7u4f18je6au/Origin2017.opj?dl=0
Screenshot:
https://www.dropbox.com/s/577gl4w06bjrut3/Origin2017.png?dl=0
|
Edited by - holyone2 on 08/28/2017 09:35:01 AM |
|
|
|
holyone2
United Kingdom
3 Posts |
Posted - 09/14/2017 : 06:23:29 AM
|
| ...? |
|
|
| |
Topic |
|
|
|
non-linear fitting: F-test