Noisy & Sparse Data

The Getting started example uses dense, noise-free simulated data. This guide applies realistic measurement noise (5% proportional CV + additive component) and sparse clinical sampling (15 PK timepoints) to verify that the IVIVC pipeline degrades gracefully under conditions typical of bioanalytical LC-MS/MS assays.

Key design choices that mirror real practice:

• Sparse PK sampling: 15 timepoints (dense around absorption, sparse in the terminal phase)
• Dissolution every 1 h: 49 timepoints (standard for automated dissolution baths)
• IV data stays clean: numerical deconvolution is sensitive to IV noise with sparse data
• Dissolution stays clean: matching real IVIVC workflows where dissolution is measured precisely in vitro

Simulated data

first(pk_df, 8)

first(diss_df, 8)

Running the pipeline

The run_ivivc call is identical to the Getting started example — the pipeline handles noisy, sparse data without any special flags.

result = run_ivivc(
    pk_df,
    diss_df;
    fast_formulation = "IV",
    validation_formulation = "FormB",
    times = collect(range(0.0, 48.0, length = 101)),
);

Model parameters

coeftable(result.model)

Plots

concentration_profiles(result)

dissolution_profiles(result)

absorption_profiles(result)

levy_plot(result)

ivivc_correlation(result)