| Author |
 Topic  |
|
|
hauli
Austria
Posts |
 Posted - 12/21/2005 : 02:27:55 AM
|
Origin Version (Select Help-->About Origin): 7.5
Operating System: windows
I`m searching for a method to test for a treatment effect in a series of matched experiments.
the whole assay is performed as follows: I got one control group and one treated. The data is fit to a dose response model (with all parameters fixed except the LOGx0, as this makes biologically sense); the whole assay is repeated twice.
So overall I got 6 values for the LOGx0 (i.e. 3 for control and 3 for treatment) and calculated the treatment effect as the difference between the LOGx0 of the control and its corresponding treatment. Unfortunately the CI for the average treatment effect calculated like this is very large and often contains the value 0.
What I want is to estimate the average treatment effect automatically and based on all of my data points (not only on the 6 values for the LOGx0)and thus based on a lot of degrees of freedom.
As I think this can be done by using global fitting models. So overall I need a model that fits all the data and provides automatically one single best fit value for the average treatment effect.
This problem sounds difficult, but I hope it is not
Thanks for any help and suggestion
HAULI
|
|
|
easwar
USA
1964 Posts |
Posted - 12/21/2005 : 3:38:24 PM
|
Hi Hauli,
If you are asking about fitting multiple datasets simultaneously/combined with one fitting equation, optionally sharing some of the fitting parameters across the datasets, and then generating one set of fit statistics such as R^2, SSR etc....that is possible with the NLSF fitting tool in Origin.
Please see the following topic in the Origin help file:
Analysis->Curve Fitting->Global Fitting of Multiple Datasets Using Shared Parameters
Easwar
OriginLab
|
 |
|
|
hauli
Austria
Posts |
Posted - 12/22/2005 : 07:40:18 AM
|
Hi easwar!
thank you for your suggestion, but what I need goes further:
as I mentioned earlier on, I got 3 control data sets with their corresponding treatment data sets. For them I can estimate (overall) 6 values for my parameter LOGx0 (3 for control data and 3 for treatment data).
What I want now, is to introduce a new parameter, the Effect. The Effect is calculated as LOGx0 of control data minus LOGx0 of the corresponding treatment.
And I want to share this new parameter Effect for all three data sets and get one single best fit value for it (with corresponding confidence intervals).
Thanks for any further suggestions and Merry Christmas!
Hauli
|
|
|
| |
Topic |
|
|
|
Global fitting