check#

This module provides functions for checking Pastas models and their components.

This can be useful to define checks that can be applied to multiple models to ensure they meet certain criteria.

Examples

Run a checklist of standard checks on a Pastas model:

checks = [
    {"func": rsq_geq_threshold, "threshold": 0.7},
    {"func": response_memory, "cutoff": 0.95, "factor_length_oseries": 0.5},
    {"func": uncertainty_gain, "n_std": 1.96},
    {"func": parameter_bounds},
]

ps.check.checklist(ml, checks)

Or use a list of checks defined in literature. Note that these checks were derived for specific questions and hydrogeological contexts and are included as inspiration, but are by no means exhaustive or applicable to all problems.

Checklists can be obtained with:

checks = ps.check.get_checks_literature("brakenhoff_2022")
ps.check.checklist(ml, checks=checks)

Checks#

The following checks built-in checks are available in this module:

  • stat_ufunc_threshold: model statistic compared to threshold using ufunc

  • rsq_geq_threshold: R^2 greater equal than threshold

  • parameter_ufunc_threshold: single parameter value compared to threshold using ufunc

  • parameters_leq_threshold: parameter values less equal than threshold

  • response_memory: memory of response functions less than fraction of calibration period

  • response_memory_vs_warmup: memory of response functions less than warmup period

  • uncertainty_gain: estimated gain less than n_std times std. deviation of the gain

  • uncertainty_parameters: estimated parameter value less than n_std times std. deviation

  • parameter_bounds: parameter optimal values not on bounds

  • acf_runs_test: check for significant autocorrelation in the noise using runs test

  • acf_stoffer_toloi_test: check for significant autocorrelation in the noise using Stoffer-Toloi test

  • acf_ljung_box_test: check for significant autocorrelation in the noise using Ljung-Box test (regular timesteps)

  • correlation_sim_vs_res: correlation between simulated and residuals less than threshold

Checklists#

The following predefined checklists are available in this module, available via the get_checks_literature() function:

  • brakenhoff_2022: checklist based on Brakenhoff et al. (2022)

  • zaadnoordijk_2019: checklist based on Zaadnoordijk et al. (2019)

Functions#

get_empty_check_dataframe()

Create an empty DataFrame for storing check results.

stat_ufunc_threshold(ml, ufunc, statistic, threshold)

Generic function to compare a model statistic with a threshold using a ufunc.

rsq_geq_threshold(ml[, threshold])

Check R^2 >= threshold.

response_memory(ml[, cutoff, factor_length_oseries, names])

Check if response function memory is shorter than fraction of calibration period.

response_memory_vs_warmup(ml[, cutoff, names])

Check if response function memory is shorter than warmup.

uncertainty_gain(ml[, n_std, names])

Check if the gain is larger than n_std times the uncertainty in the gain.

parameter_ufunc_threshold(ml, parameter, ufunc, threshold)

Generic function to compare a model statistic with a threshold using a ufunc.

parameters_leq_threshold(ml[, parameters, threshold])

Check if parameter values are less than or equal to a threshold.

parameter_bounds(ml[, parameters])

Check if the optimal parameter values are not on the lower or upper bounds.

uncertainty_parameters(ml[, parameters, n_std])

Check if parameter value is larger than n_std times the standard deviation.

guess_unit_or_dims(parameter[, return_dims])

Guess the unit or dimension of a parameter based on its name.

acf_runs_test(ml[, p_threshold])

Runs test to check if there is significant autocorrelation in the noise.

acf_stoffer_toloi_test(ml[, p_threshold])

Stoffer-Toloi test to check if there is significant autocorrelation in the noise.

acf_ljung_box_test(ml[, p_threshold])

Ljung-Box test to check if there is significant autocorrelation in the noise.

correlation_sim_vs_res(ml[, threshold])

Check if the correlation between simulated and residuals is less than threshold.

checklist(ml, checks[, report])

Run a list of checks on a Pastas model.

print_check_report(df)

Print a report of the check results.

get_checks_literature(...)

Get predefined checklists based on literature.