| Author |
 Topic  |
|
|
dgolas
USA
Posts |
 Posted - 05/13/2010 : 7:21:12 PM
|
| Is there an easy way to plot the fit of the results from a multiple linear regression as a wireframe and also show the scatter points of the data which were used to make the fit? |
|
|
larry_lan
China
Posts |
Posted - 05/14/2010 : 02:18:44 AM
|
Hi:
In Origin 8 or later version, after doing the fit (Multiple Linear Regression in your case), you can go to the MRCurve worksheet to get the fitted data and plot the curves. Now that you are asking for a wireframe plot, could you please tell me why you want to do this? As far as I know, users usually go to watch the Partial Leverage Plots and Residual Plots in Multiple Linear Regression to diagnose the fitting model. Could you please describe your problem in detail?
If you have some data to illustrate your problem, could you please email us? And then we can look into your problem.
Thanks
Larry
OriginLab Technical Services |
 |
|
|
dgolas
USA
Posts |
Posted - 05/25/2010 : 6:04:28 PM
|
Thanks for the reply and the information. I don't have any particular need for a wireframe graph, it was just the way an old graphing program we used to use displayed the fit, as you can see from the attachment.
 |
|
|
|
larry_lan
China
Posts |
Posted - 05/25/2010 : 10:21:55 PM
|
Hi:
I modified your post so everyone can see the sample graph. You can upload an image file and use the [img] tag to enclose the url.
Could you explain what's V1, V2 and V3 here in a Multiple Regression?
Thanks
Larry |
Edited by - larry_lan on 05/25/2010 10:23:57 PM |
|
|
|
dgolas
USA
Posts |
Posted - 05/26/2010 : 4:01:18 PM
|
| Sure. This was only an example, but the graph is the response of our ionization chamber (V3) for Ce-144-Pr-144 solutions at various cerium concentrations (V1, in mg/g) and densities (V2, g/mL). From your earlier post on 5/14 it looks like I have to upgrade from version 7.5. Can Version 8.1 do what you described, or do I need 8.1 Pro? |
|
|
|
easwar
USA
1964 Posts |
Posted - 05/27/2010 : 09:52:26 AM
|
Hi,
You basically want to fit two dep vars versus one indep with an equation such as
z=f(x,y)
So instead of using multiple linear regression, you could do surface fitting.
Surface fitting is available in 8.1Pro. You can then define your equation and perform the surface fit. The resulting fit-surface-data will be available in XYZ format, and you can convert that to a matrix to make a surface plot and then plot your data in the same layer to get the plot you are looking for.
Easwar
OriginLab |
|
|
|
erickzhou
17 Posts |
Posted - 07/15/2010 : 10:41:06 PM
|
Is there "multiple nonlinear regression" in origin 8?
I want to fit two dep vars verus one indep with an equation such as z=f(x,y), but my data include noise, that is, the real equation may be z=f(x,y)+noise. Do you think I can solve this problem by nonlinear surface fit? |
|
|
|
larry_lan
China
Posts |
Posted - 07/15/2010 : 11:21:34 PM
|
Hi:
The Nonlinear Curve Fitting supports multiple independent variables. Please create one fitting function in Fitting Function Organizer.
Here is a tutorial to show you how to fit multiple independent model.
Thanks
Larry |
Edited by - larry_lan on 07/15/2010 11:23:38 PM |
|
|
|
erickzhou
17 Posts |
Posted - 07/17/2010 : 11:01:54 PM
|
I create the same fitting function as "nolinear surface fitting" for Nolinear curve fitting, And I get the same result as "nolinear surface fitting". The only difference is that by "nolinear surface fitting", I get two "Fitted curves plot" instead of one "Fitted surface plot". Are these plots have some meaning?
Another question, there are some parameters in "Statistics" such as "Number of points","Degrees of Freedom","Reduced Chi-Sqr","Residual Sum of Squares","Adj. R-Square",and "Fit Status", Can you give me some explaination about these parameters?
Thank you.
|
|
|
|
erickzhou
17 Posts |
Posted - 07/17/2010 : 11:05:22 PM
|
| By the way, since I can get the same result form "nolinear surface fitting" and "nolinear surface fitting", can I say that "nolinear surface fitting" is one kind of Multiple Regression tool. |
|
|
|
larry_lan
China
Posts |
Posted - 07/18/2010 : 09:16:01 AM
|
In Nonlinear Surface Fitting, Origin outputs contour plot and you can do residual analysis directly. However, if you fit a two-independent model in Nonlinear Curve Fitting, you should make such plot by yourself. That is, you can use data outputs from Nonlinear Curve Fitting to plot the same graph as Nonlinear Surface Fitting, and vice versa.
We have explained the fitted result in our help document, and there are also help for Interpreting Regression Results.
In another post, I mentioned the basic different about linear and nonlinear fitting. What we can say is, linear model can be considered as a special case of nonlinear model.
Thanks
Larry |
|
|
| |
Topic |
|
3d plot of regression fit