4 replies to this topic

[jessicalee]

Hi, I am having some problems with the IVIVC fitting module.

 

In the results section correlation.final parameters, AbsScale and Tscale results are given. The CV% for these two parameters are given there respectively. However, according to my understanding, the coefficient of variation should be calculated from the predicted Fabs results given by the formula fit and the original input Fabs results, which in this case should only produce one CV% value, not two. Please correct me if I have misunderstood. 

 

Since the CV% is calculated from the stdError, may I ask how the stdError is calculated here?

 

Also, when I choose a correlation equation with three parameters, I find that the estimation for the first two parameters is similar to that for the equation with two parameters. Is there an order in which the different parameters are determined when performing calculations with multiple parameters? What is the strategy for determining the parameters?

 

Thanks in advance.

 

Best regards,

Jessica

 

Edited by jessicalee, 20 June 2022 - 04:30 PM.

[cradhakr]

Hi Jassica,

 

Is it possible for you to send the project details/project file to support@certara.com.

 

This would allow us to look into the project and answer your questions more precisely.

 

Thanks

Mouli

[jessicalee]

Hi Jassica,

 

Is it possible for you to send the project details/project file to support@certara.com.

 

This would allow us to look into the project and answer your questions more precisely.

 

Thanks

Mouli

Hi, Mouli

 

For the attached pictures are the results of the example for IVIVC tookit, and the project is IVIVC_Workflow.phxproj provided by Winnonlin, please see in the last post.

Looking forward to your response to the questions in the last post.

 

Big thanks,

Jessica

[Simon Davis]

Regarding your second question, with this data Tshift is not reasonable to include, note the small value with Huge CV%

That is why the other estimates did not change much, tshift is not a big contributor.

 Simon

[jessicalee]

Regarding your second question, with this data Tshift is not reasonable to include, note the small value with Huge CV%

That is why the other estimates did not change much, tshift is not a big contributor.

 Simon

 

Hi Simon,

 

I still have some questions.

 

There may be multicollinearity in Tscale and Tshift. if so, then the fitted model would be unstable (or the parameters of the model would be with a large variance). In this case, assuming that Tscale and Tshift are interactively correlated, the Tscale takes up the majority of the contribution, then the contribution of the Tshift will naturally be smaller. Are there any methods used to solve the problem of multicollinearity? (For example, the step-wise regression, simple deletion, ridge regression.)

 

In addition, the standard deviation (StdError) is calculated by subtracting what from what?

 

Big thanks,

Jessica