[HIGH]: pde convergence plots in dim > 1

[mathematicalmichael]

PDE should compute distance to projection instead of lam_true when dim > 1. In dim == 1, keep distance to truth.

• [x] plot_experiment_measurements
• [x] plot_experiment_equipment

[mathematicalmichael]

truth is pulled from the VERY messy data structure, so fixing one should fix the other automatically.

the fix should be inside of the function that calculates the statistics instead, passed to _re/_rm

tolerances:

    for _res in res:
        _prefix, _in, _rm, _re = _res
        regression_err_mean, slope_err_mean, \
            regression_err_vars, slope_err_vars, \
            sd_means, sd_vars, num_sensors = _re
        plt.plot(tolerances, regression_err_vars,
                 label=f"{_prefix:10s} slope: {slope_err_vars:1.4f}",
                 lw=linewidth)
        plt.scatter(tolerances, sd_vars, marker='x', lw=20)

...

    plt.figure(figsize=(10, 10))
    for _res in res:
        _prefix, _in, _rm, _re = _res
        regression_err_mean, slope_err_mean, \
            regression_err_vars, slope_err_vars, \
            sd_means, sd_vars, num_sensors = _re
        plt.plot(tolerances, regression_err_mean,
                 label=f"{_prefix:10s} slope: {slope_err_mean:1.4f}",
                 lw=linewidth)
        plt.scatter(tolerances, sd_means, marker='x', lw=20)

measurements:

    for _res in res:
        _prefix, _in, _rm, _re = _res
        solutions = _in[-1]
        measurements = list(solutions.keys())
        regression_mean, slope_mean, \
            regression_vars, slope_vars, \
            means, variances = _rm
        plt.plot(measurements[:len(regression_mean)], regression_mean,
                 label=f"{_prefix:4s} slope: {slope_mean:1.4f}",
                 lw=linewidth)
        plt.scatter(measurements[:len(means)], means, marker='x', lw=20)

...

    for _res in res:
        _prefix, _in, _rm, _re = _res
        regression_mean, slope_mean, \
            regression_vars, slope_vars, \
            means, variances = _rm
        plt.plot(measurements[:len(regression_vars)], regression_vars,
                 label=f"{_prefix:4s} slope: {slope_vars:1.4f}",
                 lw=linewidth)
        plt.scatter(measurements[:len(variances)], variances,
                    marker='x', lw=20)

[mathematicalmichael]

            _re = (regression_err_mean, slope_err_mean,
                   regression_err_vars, slope_err_vars,
                   sd_means, sd_vars, num_sensors)

_rm = (regression_mean, slope_mean, regression_vars, slope_vars, means, variances)

TODO:

• [x] refactor experiment_equipment to look like experiment_measurements

[mathematicalmichael]

ode will also need to change if the function isn't backwards compatible.

[mathematicalmichael]

• [ ] make sure weighted kde fails on sample-dist normal