feat: Update calculation of predictions

[mhovd]

Population and individual predictions are now calculated using the (weighted) mean/median response, instead of the (weighted) mean/median support.

[Copilot]

Pull Request Overview

This PR changes the calculation method for population and individual predictions from using the weighted mean/median support points to using the weighted mean/median of the actual response values. This represents a fundamental shift in how predictions are computed, moving from parameter-space averaging to response-space averaging.

Key changes:

• Modified prediction calculation to compute weighted mean/median of simulated responses instead of simulating from weighted mean/median parameters
• Updated weighted_median function signature to use Vec<f64> instead of Array1<f64> for consistency
• Added comprehensive simulation across all support points for each subject before aggregating results

Comments suppressed due to low confidence (4)

src/routines/output.rs:430

• The variable name spp is ambiguous. Consider renaming it to support_point or theta_row to make its purpose clearer.

            for spp in theta.row_iter() {

src/routines/output.rs:432

• The variable name spp_values is ambiguous. Consider renaming it to support_point_values or theta_values to make its purpose clearer.

                let spp_values = spp.iter().cloned().collect::<Vec<f64>>();

src/routines/output.rs:446

• The variable name outer_pred is unclear. Consider renaming it to support_point_predictions or spp_predictions to better indicate it represents predictions from a specific support point.

            for outer_pred in predictions.iter().enumerate() {

src/routines/output.rs:448

• The variable name inner_pred is unclear. Consider renaming it to time_point_prediction or prediction_at_time to better indicate it represents a prediction at a specific time point.

                for inner_pred in outer_pred.iter().enumerate() {

[Copilot]

Function parameters should use slice references (&[f64]) instead of &Vec<f64> for better API design and flexibility. This allows the function to accept both vectors and slices.

Suggested change

	fn weighted_median(data: &Vec<f64>, weights: &Vec<f64>) -> f64 {
	fn weighted_median(data: &[f64], weights: &[f64]) -> f64 {

[github-actions]

Bencher Report

Branch	157-change-how-meanmedian-population-predictions-are-calculated
Testbed	orion

Click to view all benchmark results

Benchmark	Latency	Benchmark Result seconds (s) (Result Δ%)	Upper Boundary seconds (s) (Limit %)
bimodal_ke_fit	📈 view plot 🚷 view threshold	10.13 s (-32.34%) Baseline: 14.97 s	26.46 s (38.29%)

🐰 View full continuous benchmarking report in Bencher

[mhovd]

Before this can be merged, we need to handle the occasion correctly.

output.rs