Bioequivalence and Bioavailability Forum

Hi lovemysoul!

❝ when i execute population PK using Pheonix model, CV% and standard error values for parameters are calculated and displayed in result table.

❝

❝ I think these estimated values have one on one correspondence with parameters

Don’t understand what you mean here. Can you elucidate that further?

❝ Are these calculated from estimates of each iteration?

No. Only when convergence is reached. Note that sometimes a model with the given settings, initial values, etc. has problems to converge. Always check the RetCode:

1. Relative gradient is close to zero - reported solution is likely optimal or nearly optimal.
   
2. Successive iterates within tolerance - reported solution is likely optimal or nearly optimal.
   
3. Last step failed to locate a better solution than reported solution - reported solution is likely optimal, nearly optimal, or possibly excessive nonlinearity was encountered that the optimization algorithm could not successfully deal with.

Values 4–5 (or negative ones) denote an unreliable result – or even lacking convergence. In such a case no SEs are calculated. Change the side conditions, the error model (weighting), and/or the optimization engine.

Please register at Pharsight’s Extranet and ask your questions there. Though we have some users in the forum working with Phoenix, I would expect PopPK experts to be a minority here.