Plotting

plot_truesolution(group, data::simdata, t_range, sol_group1, sol_group2; showdata=true)

Plot the ground-truth latent solution for a simulated cohort, optionally overlaying observed (simulated) measurements.

Arguments

• group: integer group selector; 1 plots sol_group1, otherwise sol_group2
• data::simdata: simulated dataset with groups group1/group2, xs, and tvals
• t_range: dense time grid (e.g., 0:0.1:10)
• sol_group1, sol_group2: matrices of shape (2, length(t_range)) with true latent trajectories
• showdata::Bool=true: overlay each subject’s observed variables as scatter

Returns

• curplot: a Plots.Plot with the chosen true solution (and optionally data scatter).

createindividualplot(m::odevae, data::simdata, idx::Int, sol::Matrix, trange, args::LossArgs; 
    title::String="", showtruesol::Bool=true, axislabs::Bool=false, 
    showOLS::Bool=true, colors_truesol::Array{String}=["#ff7f0e" "#1f77b4"])

Plot the learned latent trajectory for one simulated subject, optionally showing the true latent solution and an OLS overlay.

Arguments

• m::odevae: trained ODE-VAE model
• data::simdata: simulated dataset with xs, x_baseline, and tvals
• idx::Int: subject index in data
• sol::Matrix: true latent solution for this subject’s group (2, length(trange))
• trange: dense time grid for plotting the smooth ODE trajectory
• args::LossArgs: controls smoothing (firstonly vs. averaged multi-IC)
• title::String="": subplot title
• showtruesol::Bool=true: overlay the supplied true solution
• axislabs::Bool=false: add axis labels
• showOLS::Bool=true: overlay global OLS in latent space
• colors_truesol: two hex colors used for the true solution lines

Returns

• curplot: a Plots.Plot object combining true solution (optional), ODE-smoothed latent means, encoder means (scatter), and optional OLS fit.

createindividualplot_piecewise(m::odevae, data::simdata, ind::Int, sol::Matrix, t_range; 
    title::String="", axislabs::Bool=false, showtruesol::Bool=true,
    showOLS::Bool=true, showglobalOLS::Bool=false,
    colors_truesol::Array{String}=["#ff7f0e" "#1f77b4"])

Plot piecewise local ODE solutions between observed time points for one simulated subject, optionally overlaying the group’s true solution, per-segment OLS, and a global OLS fit.

Arguments

• m::odevae: trained ODE-VAE model
• data::simdata: simulated dataset with xs, x_baseline, and tvals
• ind::Int: subject index in data
• sol::Matrix: true solution (2, length(t_range)) for the subject’s group
• t_range: dense time grid for plotting the true solution
• title::String="": subplot title
• axislabs::Bool=false: add axis labels
• showtruesol::Bool=true: overlay true solution
• showOLS::Bool=true: overlay OLS per segment aligned to each segment start
• showglobalOLS::Bool=false: overlay an OLS fit over the full observed window
• colors_truesol: two hex colors for the true solution lines

Returns

• curplot: a Plots.Plot object with piecewise ODE curves, uncertainty ribbons, and optional OLS overlays.

eval_z_trajectories(m::odevae, data::simdata, inds::Array{Int}, 
    sol_group1::Matrix, sol_group2::Matrix, t_range, args::LossArgs; 
    title::String="", showtruesol::Bool=true, axislabs::Bool=false, 
    showOLS::Bool=true, swapcolorcoding::Bool=false)

Display a panel of per-subject latent trajectory plots for simulated data, selecting the appropriate ground-truth solution by group membership.

Arguments

• m::odevae: trained ODE-VAE model
• data::simdata: simulated dataset with group1, group2, xs, and tvals
• inds::Array{Int}: subject indices to visualize
• sol_group1, sol_group2: true solutions (2, length(t_range)) for each group
• t_range: dense time grid for smooth curves
• args::LossArgs: smoothing controls (firstonly, etc.)
• title::String="": panel title
• showtruesol::Bool=true: overlay true solutions
• axislabs::Bool=false: add axis labels in subplots
• showOLS::Bool=true: overlay global OLS per subject
• swapcolorcoding::Bool=false: swap the two colors used for the true solution

Returns

Nothing. Displays a Plots.Plot panel to the current display.

createindividualplot(m::odevae, testdata::SMATestData, args::LossArgs, patient_id; 
    axislabs::Bool=false, title::String="", showOLS::Bool=true)

Plot the learned latent trajectory for a single SMArtCARE patient, comparing the ODE-smoothed latent mean to encoder means (and optionally an OLS fit).

The smooth trajectory is obtained either from a single-IC rollout (args.firstonly == true) or by averaging predictions from all time points as initial conditions.

Arguments

• m::odevae: trained ODE-VAE model
• testdata::SMATestData: dataset holding time series, baselines, ids, and time points
• args::LossArgs: controls the smoothing strategy (e.g., firstonly)
• patient_id: identifier present in testdata.ids
• axislabs::Bool=false: add axis labels (“time in months”, “value of latent representation”)
• title::String="": subplot title
• showOLS::Bool=true: overlay a (global) piecewise-aligned linear regression in latent space

Returns

• curplot: a Plots.Plot object with:
  • smooth ODE latent means over a dense time grid,
  • encoder latent means as scatter,
  • optional OLS line(s).

createindividualplot_piecewise(m::odevae, testdata::SMATestData, patient_id; 
    title::String="", showOLS::Bool=true, axislabs::Bool=false)

Plot local, piecewise ODE solutions between consecutive observed time points for one SMArtCARE patient, with uncertainty ribbons from the encoder’s log-variance and optional OLS overlays.

Arguments

• m::odevae: trained ODE-VAE model
• testdata::SMATestData: dataset holding time series, baselines, ids, and time points
• patient_id: identifier present in testdata.ids
• title::String="": subplot title
• showOLS::Bool=true: overlay per-segment linear regressions aligned at each segment start
• axislabs::Bool=false: add axis labels

Returns

• curplot: a Plots.Plot object with segment-wise ODE rolls, encoder means (scatter), and optional OLS fits.

plot_selected_ids(m::odevae, testdata::SMATestData, args::LossArgs, selected_ids::Array; 
    showOLS::Bool=true, layout=nothing, size=nothing, 
    save_plot::Bool=false, save_path::String="")

Create a panel of global ODE-smoothed latent trajectories for multiple SMArtCARE patients. Uses the same smoothing strategy as createindividualplot.

Arguments

• m::odevae: trained ODE-VAE model
• testdata::SMATestData: dataset with time series, baselines, ids, and time points
• args::LossArgs: controls smoothing (e.g., firstonly)
• selected_ids::Array: collection of patient IDs to plot
• showOLS::Bool=true: show global OLS overlays
• layout=nothing: layout tuple (rows, cols); defaults to (length(selected_ids)÷4, 4)
• size=nothing: (width, height) in pixels; defaults to (1200, round(200/3)*length(selected_ids))
• save_plot::Bool=false: save figure to disk
• save_path::String="": path passed to savefig if save_plot == true

Returns

• panelplot: a Plots.Plot panel with the requested subplots.

Notes

Saves the panel to save_path when save_plot=true.

plot_selected_ids_piecewise(m::odevae, testdata::SMATestData, selected_ids::Array; 
    showOLS::Bool=true, layout=nothing, size=nothing, 
    save_plot::Bool=false, save_path::String="")

Create a panel of piecewise ODE plots for multiple SMArtCARE patients.

Arguments

• m::odevae: trained ODE-VAE model
• testdata::SMATestData: dataset with time series, baselines, ids, and time points
• selected_ids::Array: collection of patient IDs to plot (order preserved)
• showOLS::Bool=true: show per-segment OLS overlays in each subplot
• layout=nothing: layout tuple (rows, cols); defaults to (length(selected_ids)÷4, 4)
• size=nothing: (width, height) in pixels; defaults to (1200, round(200/3)*length(selected_ids))
• save_plot::Bool=false: save figure to disk
• save_path::String="": path passed to savefig if save_plot == true

Returns

• panelplot: a Plots.Plot panel with the requested subplots.

Notes

Saves the panel to save_path when save_plot=true.