4.9. IM Comparison

imselection.compare_ims(im_results_dict, analysis_type='MCA', metric='all', reference_im_key=None, damage_threshold_index=0)[source]

Compare N ≥ 2 intensity measures by efficiency, proficiency, practicality, and RSM.

Parameters:

  • im_results_dict (dict) – Mapping {im_name: cloud_dict} (for MCA) or {im_name: ida_dict} (for IDA).

  • analysis_type ({'MCA', 'IDA'}) – Which NDAP was used to generate the results.

  • metric ({'all', 'efficiency', 'proficiency', 'practicality', 'rsm'}) – Which metrics to compute. 'all' computes all four.

  • reference_im_key (str or None) – IM name to use as IM₁ in RSM pairwise comparisons. If None, the first key in im_results_dict is used.

  • damage_threshold_index (int) – Damage-state index for proficiency computation. Default 0.

Returns:

  • dict with keys

  • * 'ranking' — pd.DataFrame with columns im_name, – efficiency, proficiency, practicality, rsm_vs_reference, rank_efficiency, rank_proficiency, rank_practicality, rank_rsm

  • * 'rsm_matrix' — dict[str, dict[str, float]] where – rsm_matrix[im_i][im_j] = RSM(IM_j vs IM_i)

  • * 'metric' — the metric argument used

  • * 'analysis_type' — the analysis_type argument used

  • * 'reference_im_key' — the IM used as IM₁ for RSM

Raises:

ValueError – If fewer than 2 IMs are supplied or analysis_type is invalid.

Example

from openquake.vmtk.imselection import imselection

ims = imselection()
# cloud_pga, cloud_sa, cloud_avgsa: outputs of postprocessor.process_mca_results()
results = {
    "PGA":    cloud_pga,
    "Sa(T1)": cloud_sa,
    "AvgSa":  cloud_avgsa,
}
ranking = ims.compare_ims(results, analysis_type="MCA", metric="all")
print(ranking["ranking"])