| T O P I C R E V I E W |
| giordandue |
Posted - 12/12/2018 : 03:27:18 AM
Origin Ver. and Service Release (Select Help-->About Origin): Origin 2016 sr2
Operating System:Windows 10
Hello,
is it possible to calculate the correlation coefficient of a given (linear) function? In the specific case, I have data point and I would calculate how well the simple function y=0.5x correlates the data.
Thank you in advance |
| 8 L A T E S T R E P L I E S (Newest First) |
| giordandue |
Posted - 12/14/2018 : 09:28:45 AM
Thank you a lot James!
Greetings from Italy
Giordano |
| YimingChen |
Posted - 12/14/2018 : 09:11:17 AM
Hi Giordano,
So basically the purpose of regression is to use a model to explain the variation of the data. R^2 is the ratio of explained variation to the total variation of the data. If you look at eqn 21 on this page for the definition of R^2,
https://www.originlab.com/doc/Origin-Help/NLFit-Theory#R-Square_.28COD.29
a negative R^2 means the residual variation (RSS) is even larger than the total variation (TSS) , which means the model doesn't help in explaining your data. Hope this answers your question. Thank you.
James
|
| giordandue |
Posted - 12/14/2018 : 01:27:37 AM
Thank you very much James..
It seems to works, but I have to understand the meaning of the resulting negative values of both R-square and Adjusted R-square and how to manage it...
I add the figure of the fitting result, for the function y=0.5x I have forced:
Kind Regards
Giordano |
| YimingChen |
Posted - 12/13/2018 : 10:46:45 AM
Hi,
You could use Nonlinear Curve fitting tool to achieve that. In NLFit dialog, you can change the parameters of the fitting function. Then instead of performing a fitting, just click Done button, it outputs the R^2 with current parameter settings. This way you can update your function and check it's correlation with your data.
Hope this works for you. Thanks.
James |
| giordandue |
Posted - 12/13/2018 : 09:15:16 AM
Hi Jaqueline,
your suggestion appears as quite complicate and requiring tools that I have not available.
Is there not other available method to check how well a defined function fits data?
Regards |
| JacquelineHe |
Posted - 12/13/2018 : 05:11:40 AM
Hi,
Maybe you can try to use the function to create a dataset (Set Column Value tool), and then use Correlation Coefficient tool to analysis this function datasets and your original dataset.
Do you think the method is good for your?
About Set Column Value:
https://www.originlab.com/doc/Tutorials/SetColVal
About Correlation Coefficient tool:
https://www.originlab.com/doc/Origin-Help/Correlation-Coefficient
Thanks
Jacqueline
OriginLab |
| giordandue |
Posted - 12/13/2018 : 03:01:49 AM
Hi Jaqueline,
I think you have not understood my question.
I do not want to do a linear fitting of the data, but, if possible, to calcolate the correlation coefficient for a given function (in the case Y=0.5X). In other word, let we assume that Y=0.3X is the best linear fitting of the data with R2=0.8: I would calculate R2 for Y=0.5X.
Thank you in advance |
| JacquelineHe |
Posted - 12/12/2018 : 05:39:39 AM
Hi,
If you do the Linear Fit ("Analysis: Fitting: Linear Fit" menu) for your datasets, you will get the "Pearson's r" (Pearson correlation coefficient) in the report sheet.
Thanks
Originlab
Jacqueline
|
