Example 7 — Full Diagnostic Plot Gallery

Script: examples/07_diagnostic_plots.py

A complete walkthrough of every diagnostic plot function using the theophylline FO model from Example 1 — Theophylline (FO).

Output

openpkpd/estimation/foce.py:236: UserWarning: ETA1 shrinkage is 72.5% (>30%). EBE-based analyses for this parameter may be unreliable.
  res.compute_shrinkage()
openpkpd/estimation/foce.py:236: UserWarning: ETA2 shrinkage is 73.6% (>30%). EBE-based analyses for this parameter may be unreliable.
  res.compute_shrinkage()
openpkpd/estimation/foce.py:236: UserWarning: ETA3 shrinkage is 82.3% (>30%). EBE-based analyses for this parameter may be unreliable.
  res.compute_shrinkage()
Running FOCE...
Method: FOCEI
OFV: 278.7017
AIC: 298.7017
BIC: inf   (n_obs = 0)
n_parameters: 10
Converged: False
THETA: [ 1.5   0.08 30.  ]
OMEGA (diagonal): [0.5 0.3 0.3]
SIGMA (diagonal): [0.1]
ETA shrinkage: ['72.5%', '73.6%', '82.3%']
Shrinkage warnings:
  ETA1 shrinkage is 72.5% (>30%). EBE-based analyses for this parameter may be unreliable.
  ETA2 shrinkage is 73.6% (>30%). EBE-based analyses for this parameter may be unreliable.
  ETA3 shrinkage is 82.3% (>30%). EBE-based analyses for this parameter may be unreliable.

Created 14 figures.
Figures saved to docs/_static/examples

Figures

GOF panel (2×3 composite)

The panel contains DV vs IPRED, DV vs PRED, CWRES vs TIME, CWRES vs PRED, CWRES Q-Q, and |IWRES| vs IPRED.

GOF panel

Individual GOF plots

DV vs IPRED DV vs PRED CWRES vs TIME CWRES vs PRED CWRES Q-Q |IWRES| vs IPRED

PK concentration-time plots

Spaghetti plot Concentration-time Mean profile

ETA diagnostics

ETA histograms ETA pairs ETA1 vs weight

OFV convergence

OFV history

Interpreting GOF plots

Plot	What to look for
DV vs IPRED / PRED	Points close to the identity line, no systematic bias
CWRES vs TIME	Random scatter around zero; no trend or funnel
CWRES vs PRED	Random scatter; no heteroscedasticity
CWRES Q-Q	Points on the diagonal line; normality of residuals
\|IWRES\| vs IPRED	Homogeneous spread; no increasing trend (proportional error check)
ETA histograms	Symmetric, approximately normal distributions
ETA pairs	No strong correlations (would suggest OMEGA mis-specification)
ETA vs covariates	Flat relationship (no unexplained covariate effect)