[Temporal fit] add gmle fit method by default for non stationary trends

experiment/paper_past_snow_loads/check_mle_convergence_for_trends/study_visualizer_for_shape_repartition.py

@@ -10,18 +10,8 @@ from experiment.trend_analysis.abstract_score import MeanScore
 
 class StudyVisualizerForShape(StudyVisualizerForNonStationaryTrends):
 
-    def __init__(self, study: AbstractStudy, show=True, save_to_file=False, only_one_graph=False, only_first_row=False,
-                 vertical_kde_plot=False, year_for_kde_plot=None, plot_block_maxima_quantiles=False,
-                 temporal_non_stationarity=False, transformation_class=None, verbose=False, multiprocessing=False,
-                 complete_non_stationary_trend_analysis=False, normalization_under_one_observations=True,
-                 score_class=MeanScore, uncertainty_methods=None, non_stationary_contexts=None,
-                 uncertainty_massif_names=None, effective_temporal_covariate=2017, relative_change_trend_plot=True):
-        super().__init__(study, show, save_to_file, only_one_graph, only_first_row, vertical_kde_plot,
-                         year_for_kde_plot, plot_block_maxima_quantiles, temporal_non_stationarity,
-                         transformation_class, verbose, multiprocessing, complete_non_stationary_trend_analysis,
-                         normalization_under_one_observations, score_class, uncertainty_methods,
-                         non_stationary_contexts, uncertainty_massif_names, effective_temporal_covariate,
-                         relative_change_trend_plot)
+    def __init__(self, study: AbstractStudy, **kwargs):
+        super().__init__(study, **kwargs)
 
     @cached_property
     def massif_name_to_unconstrained_shape_parameter(self):

experiment/paper_past_snow_loads/paper_main_utils.py

@@ -2,15 +2,18 @@ from collections import OrderedDict
 
 from experiment.paper_past_snow_loads.study_visualizer_for_non_stationary_trends import \
     StudyVisualizerForNonStationaryTrends
+from extreme_fit.model.margin_model.linear_margin_model.abstract_temporal_linear_margin_model import \
+    TemporalMarginFitMethod
 
 
 def load_altitude_to_visualizer(altitudes, massif_names, non_stationary_uncertainty, study_class, uncertainty_methods,
                                 study_visualizer_class=StudyVisualizerForNonStationaryTrends,
                                 save_to_file=True):
+    fit_method = TemporalMarginFitMethod.extremes_fevd_gmle
     altitude_to_visualizer = OrderedDict()
     for altitude in altitudes:
         altitude_to_visualizer[altitude] = study_visualizer_class(
             study=study_class(altitude=altitude), multiprocessing=True, save_to_file=save_to_file,
             uncertainty_massif_names=massif_names, uncertainty_methods=uncertainty_methods,
-            non_stationary_contexts=non_stationary_uncertainty)
+            non_stationary_contexts=non_stationary_uncertainty, fit_method=fit_method)
     return altitude_to_visualizer

experiment/paper_past_snow_loads/result_trends_and_return_levels/main_result_trends_and_return_levels.py

@@ -83,8 +83,8 @@ def major_result():
     uncertainty_methods = [ConfidenceIntervalMethodFromExtremes.my_bayes,
                            ConfidenceIntervalMethodFromExtremes.ci_mle][1:]
     massif_names = None
-    study_classes = paper_study_classes[:2]
-    study_classes = [CrocusSnowLoad3Days, CrocusSnowLoad5Days, CrocusSnowLoad7Days][::-1]
+    study_classes = paper_study_classes[:1]
+    # study_classes = [CrocusSnowLoad3Days, CrocusSnowLoad5Days, CrocusSnowLoad7Days][::-1]
     for study_class in study_classes:
         if study_class == CrocusSnowLoadEurocode:
             non_stationary_uncertainty = [False]

experiment/paper_past_snow_loads/study_visualizer_for_non_stationary_trends.py

@@ -2,7 +2,7 @@ from collections import OrderedDict
 
 import matplotlib.pyplot as plt
 from multiprocessing.pool import Pool
-from typing import Dict
+from typing import Dict, List
 
 import numpy as np
 from cached_property import cached_property
@@ -48,12 +48,14 @@ class StudyVisualizerForNonStationaryTrends(StudyVisualizer):
                  uncertainty_massif_names=None,
                  effective_temporal_covariate=2017,
                  relative_change_trend_plot=True,
-                 non_stationary_trend_test_to_marker=None):
+                 non_stationary_trend_test_to_marker=None,
+                 fit_method=None):
         super().__init__(study, show, save_to_file, only_one_graph, only_first_row, vertical_kde_plot,
                          year_for_kde_plot, plot_block_maxima_quantiles, temporal_non_stationarity,
                          transformation_class, verbose, multiprocessing, complete_non_stationary_trend_analysis,
                          normalization_under_one_observations, score_class)
         # Add some attributes
+        self.fit_method = fit_method
         self.non_stationary_trend_test_to_marker = non_stationary_trend_test_to_marker
         self.relative_change_trend_plot = relative_change_trend_plot
         self.effective_temporal_covariate = effective_temporal_covariate
@@ -118,7 +120,12 @@ class StudyVisualizerForNonStationaryTrends(StudyVisualizer):
             quantile_level = self.massif_name_to_eurocode_quantile_level_in_practice[massif_name]
             non_stationary_trend_test = [
                 t(years=x, maxima=y, starting_year=starting_year, quantile_level=quantile_level)
-                for t in self.non_stationary_trend_test]
+                for t in self.non_stationary_trend_test]  # type: List[AbstractGevTrendTest]
+            # Set appropriate fit method for all objects (#todo: set this parameter directly in the init function)
+            if self.fit_method is not None:
+                for t in non_stationary_trend_test:
+                    t.fit_method = self.fit_method
+            # Extract
             trend_test_that_minimized_aic = sorted(non_stationary_trend_test, key=lambda t: t.aic)[0]
             massif_name_to_trend_test_that_minimized_aic[massif_name] = trend_test_that_minimized_aic
         return massif_name_to_trend_test_that_minimized_aic
@@ -131,7 +138,10 @@ class StudyVisualizerForNonStationaryTrends(StudyVisualizer):
 
     @cached_property
     def _max_abs_change(self):
-        return self.global_max_abs_change if self.global_max_abs_change is not None else self.max_abs_change
+        max_abs_change = self.global_max_abs_change if self.global_max_abs_change is not None else self.max_abs_change
+        if max_abs_change == 0:
+            max_abs_change = 1e-10
+        return max_abs_change
 
     def plot_trends(self, max_abs_tdrl=None,  add_colorbar=True):
         if max_abs_tdrl is not None: