Question of output for individual modelling result

Phoenix Certara Forums Modelling and Simulation
1

Dear all,

I would like to confirm the following issue:

  1. How did the IWRES be calculated from the output data? Is the IWRES=(DV-predicted)/weight ?

  2. what is detail calculation information of LOESS regression including (1) degree of local polynomials, (2) smoothing parameter, and (3) weight function in the IWRES vs IPRED (or I****WRES vs IVAR )?

Thanks in Advance

Emily

2

see equation 16 in my paper for IWRES definition this is a standard residual used in the pmx work
https://ascpt.onlinelibrary.wiley.com/doi/pdf/10.1002/psp4.12161

the developers can comment on what is being used in the Phoenix software as for the loess you cannot control anything about it except the n of points.

developers can disclose what is being used under the hood ?

3

Hi Emily,

the Phoenix Loess function uses the Cleveland-Grosse Bell labs fortran 77 implementation in Netlib. See http://www.netlib.org/a/loess for a description. The smoothing parameter, alpha, is determined automatically and depends on the number of points: if(n.gt.10) alpha=1 if(n.gt.10) alpha=.9 if(n.gt.15) alpha=.8 if(n.gt.20) alpha=.7 if(n.gt.50) alpha=.6 if(n.ge.100) alpha=.5 if(n.ge.200) alpha=.5 if(n.ge.500) alpha=.5 if(n.ge.1000) alpha=0.5 In most of the case, the default smoothing value is good enough.

Bernd

4

in the future version of Phoenix you will be able to have more control on thow the smoothing alpha is set

5

Dear all,

Thanks for all of your reply. Based on the article and link, I also have some questions:

  1. IWRES- I would like to confirm If the weight for an additive error model under individual mode is the square root of the variance of the residual error? cause we are NOT using population PK model but individual mode.

  2. The weight of LOESS created from Phoenix is described in the code written in Fortran. However, I am not able to get understand with it. Could anyone help me to transfer into a layman’s term how the weight is calculated ?

Thank you very much

Emily

6

https://onlinehelp.certara.com/phoenix/8.2/index.html?#t=topics%2Fnlmeplotoutput.htm&rhsearch=IWRES&rhhlterm=IWRES&rhsyns=%20

here you have the online help definition of IWRES:

WRES=(DV – IPRED)/STD (1)

(see also the definition for STD for additive and other error models)

the loess “span” depends on the n of neighbor points not sure I understand the “Weight” of loess can you clarify ?

7

Dear sir,

For the weight function in LOESS, it is to give individual observations in the sum of squared residuals a different weight in the local fitting criterion.

thank you very much

Emily