[projection snowfall] fix error in one fold fit for the temperature covariate...

[projection snowfall] fix error in one fold fit for the temperature covariate & for the significance.
modify number of parameters for the spline, and only counts the number of coefficients and not the number of knots.

extreme_fit/function/param_function/spline_coef.py

@@ -29,7 +29,8 @@ class SplineCoef(AbstractCoef):
 
     @property
     def nb_params(self):
-        return self.nb_knots + self.nb_coefficients
+        # return self.nb_knots + self.nb_coefficients
+        return self.nb_coefficients
 
 class SplineAllCoef(LinearCoef):
 

extreme_fit/model/result_from_model_fit/result_from_extremes/result_from_evgam.py

@@ -53,7 +53,8 @@ class ResultFromEvgam(AbstractResultFromExtremes):
 
     @property
     def nb_parameters(self):
-        return len(np.array(self.name_to_value['coefficients']))  + self.nb_knots
+        # return len(np.array(self.name_to_value['coefficients']))  + self.nb_knots
+        return len(np.array(self.name_to_value['coefficients']))
 
     @property
     def aic(self):

extreme_trend/ensemble_fit/visualizer_for_projection_ensemble.py

@@ -6,6 +6,8 @@ from extreme_data.meteo_france_data.adamont_data.adamont_gcm_rcm_couples import
 from extreme_data.meteo_france_data.adamont_data.adamont_scenario import gcm_rcm_couple_to_str
 from extreme_data.meteo_france_data.scm_models_data.abstract_study import AbstractStudy
 from extreme_fit.distribution.gev.gev_params import GevParams
+from extreme_fit.model.margin_model.linear_margin_model.abstract_temporal_linear_margin_model import \
+    AbstractTemporalLinearMarginModel
 from extreme_fit.model.margin_model.polynomial_margin_model.spatio_temporal_polynomial_model import \
     AbstractSpatioTemporalPolynomialModel
 from extreme_fit.model.margin_model.utils import MarginFitMethod
@@ -25,7 +27,7 @@ from projects.projected_extreme_snowfall.results.plot_relative_change_in_return_
 class VisualizerForProjectionEnsemble(object):
 
     def __init__(self, altitudes_list, gcm_rcm_couples, study_class, season, scenario,
-                 model_classes: List[AbstractSpatioTemporalPolynomialModel],
+                 model_classes: List[AbstractTemporalLinearMarginModel],
                  ensemble_fit_classes=None,
                  massif_names=None,
                  fit_method=MarginFitMethod.extremes_fevd_mle,

extreme_trend/one_fold_fit/one_fold_fit.py

@@ -13,6 +13,7 @@ from extreme_fit.distribution.gev.gev_params import GevParams
 from extreme_fit.distribution.gumbel.gumbel_gof import goodness_of_fit_anderson
 from extreme_fit.estimator.margin_estimator.utils import fitted_linear_margin_estimator_short
 from extreme_fit.function.param_function.polynomial_coef import PolynomialAllCoef, PolynomialCoef
+from extreme_fit.model.margin_model.linear_margin_model.temporal_linear_margin_models import StationaryTemporalModel
 from extreme_fit.model.margin_model.polynomial_margin_model.gev_altitudinal_models import StationaryAltitudinal
 from extreme_fit.model.margin_model.polynomial_margin_model.gev_altitudinal_models_only_altitude_and_scale import \
     AltitudinalOnlyScale, StationaryAltitudinalOnlyScale
@@ -209,7 +210,7 @@ class OneFoldFit(object):
         return coordinate
 
     def _compute_shape_for_reference_altitude(self, estimator):
-        coordinate = self.get_coordinate(self.altitude_plot, self.last_year)
+        coordinate = self.get_coordinate(self.altitude_plot, self.covariate_after)
         gev_params = estimator.function_from_fit.get_params(coordinate, is_transformed=False)
         shape = gev_params.shape
         return shape
@@ -296,7 +297,10 @@ class OneFoldFit(object):
         elif isinstance(self.best_estimator.margin_model, AltitudinalShapeLinearTimeStationary):
             return self.model_class_to_estimator_with_finite_aic[AltitudinalShapeLinearTimeStationary]
         else:
-            return self.model_class_to_estimator_with_finite_aic[StationaryAltitudinal]
+            if isinstance(self.altitude_group, DefaultAltitudeGroup):
+                return self.model_class_to_estimator_with_finite_aic[StationaryTemporalModel]
+            else:
+                return self.model_class_to_estimator_with_finite_aic[StationaryAltitudinal]
 
     @property
     def likelihood_ratio(self):

projects/projected_extreme_snowfall/results/main_projections_ensemble.py

@@ -50,13 +50,13 @@ def main():
     AbstractExtractEurocodeReturnLevel.ALPHA_CONFIDENCE_INTERVAL_UNCERTAINTY = 0.2
     scenarios = [AdamontScenario.rcp85_extended]
 
-    fast = True
+    fast = None
     for scenario in scenarios:
         gcm_rcm_couples = get_gcm_rcm_couples(scenario)
         if fast is None:
-            gcm_rcm_couples = gcm_rcm_couples[:2]
+            gcm_rcm_couples = gcm_rcm_couples[:]
             AbstractExtractEurocodeReturnLevel.NB_BOOTSTRAP = 10
-            altitudes_list = [1800, 2100]
+            altitudes_list = [900, 1800, 2700, 3600]
         elif fast:
             gcm_rcm_couples = gcm_rcm_couples[:2]
             AbstractExtractEurocodeReturnLevel.NB_BOOTSTRAP = 10
@@ -81,7 +81,6 @@ def main():
         massif_names = ['Vanoise']
         model_classes = SPLINE_MODELS_FOR_PROJECTION_ONE_ALTITUDE
 
-
         visualizer = VisualizerForProjectionEnsemble(
             altitudes_list, gcm_rcm_couples, study_class, Season.annual, scenario,
             model_classes=model_classes,

projects/projected_extreme_snowfall/results/plot_relative_change_in_return_level.py

@@ -5,6 +5,7 @@ import numpy as np
 
 from extreme_trend.one_fold_fit.altitudes_studies_visualizer_for_non_stationary_models import \
     AltitudesStudiesVisualizerForNonStationaryModels
+from root_utils import get_display_name_from_object_type
 from spatio_temporal_dataset.coordinates.temporal_coordinates.temperature_covariate import \
     AnomalyTemperatureWithSplineTemporalCovariate
 
@@ -31,8 +32,10 @@ def plot_relative_dynamic_in_return_level(massif_names, visualizer_list: List[
 
 
 def plot_curve(ax, massif_name, visualizer: AltitudesStudiesVisualizerForNonStationaryModels):
-    temperatures_list = np.linspace(1, 4, num=4)
+    temperatures_list = np.linspace(1, 5, num=40)
     one_fold_fit = visualizer.massif_name_to_one_fold_fit[massif_name]
+    print(get_display_name_from_object_type(type(one_fold_fit.best_margin_model)),
+          "significant={}".format(one_fold_fit.is_significant))
     return_levels = [one_fold_fit.relative_changes_of_moment([None], order=None,
                                                              covariate_before=1,
                                                              covariate_after=t)[0] for t in temperatures_list]

test/test_extreme_fit/test_estimator/test_temporal_estimator/test_gev_temporal_spline.py

@@ -52,7 +52,6 @@ class TestGevTemporalSpline(unittest.TestCase):
                                                    starting_year=None,
                                                    fit_method=self.fit_method)
         # Checks that parameters returned are indeed different
-        print(self.start_year, self.middle_year, self.last_year)
         mle_params_estimated_year_first = estimator.function_from_fit.get_params(np.array([self.start_year])).to_dict()
         mle_params_estimated_year_middle = estimator.function_from_fit.get_params(np.array([self.middle_year])).to_dict()
         mle_params_estimated_year_last = estimator.function_from_fit.get_params(np.array([self.last_year])).to_dict()