Instructor's Note

Start with the 01-population.jl file. Show users how to load Pumas with the using statement. Load an NMTRAN-formatted DataFrame, and show a preview of the data by going over the column names. Explain the difference between evids and why some rows have missing values (measurement rows, evid == 1). We suggest using the iv_sd_3 dataset from PharmaDatasets.jl.

Once users are familiar with the NMTRAN format dataset, introduce the read_pumas function to parse NMTRAN-formatted DataFrames into a Population. Go over the read_pumas' docstring (?read_pumas) with extra attention on the keyword arguments. The keyword arguments that you should focus on are:

• observations
• covariates
• id
• time
• evid
• amt
• cmt

If users need to parse complex dosing datasets, e.g. datasets with II, ADDL, SS, RATE, or ROUTE; explain the following read_pumas's keyword arguments:

• ii
• addl
• ss
• rate
• route

Additionally, if users need to deal with datasets that have a missing observation column, i.e. MDV in NMTRAN-formatted datasets; explain the mdv read_pumas's keyword arguments.

Proceed by parsing the NMTRAN-formatted dataset into a Population. Explain that Population is simply a collection (Vector) of Subjects by indexing it and showing the Subject result. Like any Julia Vector you can also slice a Population. Show how to slice a Population into a subset of the original Population.

Showcase how to do the opposite, convert a Population or a Subject into an NMTRAN-formatted DataFrame with the DataFrame constructor, e.g. DataFrame(pop). Additionally, demonstrate that you can use the DataFrame constructor into any of the previous slices and indexes of the original Population.

Move to the 02-model.jl. Start by explaining the @model macro: it allows you to specify model blocks inside it. With respect to the model blocks, begin with the @metadata block and stress the importance of specifying model description and time units. Explain the @param block with a focus on the different domains, e.g. RealDomain and PDiagDomain. Don't forget to teach users how to type LaTeX symbols in Julia/Pumas. Explain the @random block with a focus on the probabilistic assignment ~. Explain the @covariates block and make sure that users understand that the covariates need to be also included in the read_pumas function when parsing the data into a Population. Explain the @pre block making analogies to NONMEM's $PK model block. Explain the @dynamics block by showing examples of both analytical solutions and systems of ordinary differential equations, i.e. Central1 versus Central' = -(CL/VC) * Central. Explain the @derived block with a focus on deterministic = and probabilistic ~ assignments, also for the DVs in this block remark users that they should be included in the read_pumas function as values to the observations keyword argument.

Proceed to the 03-fit.jl. Begin by showing how to define a initial parameter values NamedTuple. Then, highlight that the fit function takes four positional arguments:

1. model: which model to fit
2. population: which population to fit
3. initial_parameters: a NamedTuple of initial parameter estimates
4. estimation_method: which estimation method to use; for maximum likelihood: FOCE, NaivePooled, and LaplaceI are the most common

Perform a fit using FOCE(). Perform a fit using NaivePooled. Don't forget to explain the omegas keyword argument when fitting with NaivePooled. Perform a fit using LaplaceI. Perform a fit with fixed parameter values using the constantcoef keyword argument. Show how to get the parameter estimates using coef for a NamedTuple, and coeftable for a DataFrame. Finally, show how to get a DataFrame of the individual subject-specific parameters with icoef.

Open 04-infer.jl. Showcase how to calculate confidence intervals (CIs) using the infer function. The default case will generate CIs using the variance-covariance matrix calculated using the sandwich estimator. You can pass an optional second positional argument for alternate ways to generate CIs. If you pass Pumas.Bootstrap() as the second positional argument you will generate CIs using bootstrap, which by default fits the model to 200 bootstrapped samples. If you pass Pumas.SIR() as the second positional argument you will generate CIs using the sampling importance resampling (SIR) method. Warn the user that Pumas.SIR() does not have default values, hence it is necessary to always specify the keyword arguments samples and resamples.

Finalize the workshop by asking for questions and getting feedback from the users.

Get in touch

If you have any suggestions or want to get in touch with our education team, please send an email to training@pumas.ai.

License

This content is licensed under Creative Commons Attribution-ShareAlike 4.0 International.