def default_arg(arg=None): ##{{{
"""
Dictionary of arguments of __init__
Parameters
----------
arg : None or dict
dictionary of arguments already fixed
Returns
-------
default: dict
Arguments of __init__, elements of "arg" are kept
"""
default = {
"tails": "upper",
"link_scale": sdt.IdLink(),
"link_shape": sdt.IdLink(),
"method": "MLE",
"verbose": False
}
if arg is not None:
for key in arg:
default[key] = arg[key]
return default
def to_netcdf( self ):##{{{
ncargs = { "ns_law_name" : "GEVPr" }
ncargs[ "ns_law_param_loc_cst" ] = False
ncargs[ "ns_law_param_loc_link" ] = sdt.IdLink()
ncargs[ "ns_law_param_scale_cst" ] = False
ncargs[ "ns_law_param_scale_link" ] = sdt.IdLink()
ncargs[ "ns_law_param_shape_cst" ] = False
ncargs[ "ns_law_param_shape_link" ] = sdt.IdLink()
ncargs[ "ns_law_param_alpha_cst" ] = False
ncargs[ "ns_law_param_alpha_link" ] = sdt.IdLink()
return ncargs
def __init__(self,
tails="upper",
link_loc=sdt.IdLink(),
link_scale=sdt.ExpLink(),
link_shape=sdt.LogitLink(-0.5, 0.5),
method="MLE",
no_test=True,
verbose=False): ##{{{
"""
"""
NSAbstractModel.__init__(self)
self._tails = 1 if tails == "upper" else -1
self._gev = sd.GEVLaw(method=method)
self._link = {
"loc": link_loc,
"scale": link_scale,
"shape": link_shape
}
self._no_test = no_test
self._verbose = verbose
self._loc0 = None
self._loc1 = None
self._scale0 = None
self._scale1 = None
self._shape = None
self.loc = None
self.scale = None
self.shape = None
self._loct = None
self._scalet = None
self._shapet = None
def __init__( self , name , is_cst , link , name_tex = None ):##{{{
self.name = name
self.link = link if link is not None else sdt.IdLink()
self.is_cst = is_cst
self.n_params = 1 if self.is_cst else 2
self.coef_ = None
self._paramst = None
self.name_tex = name_tex
if name_tex is None:
self.name_tex = "\mathrm{" + self.name + "}"
def __init__(self,
link_loc=sdt.IdLink(),
link_scale=sdt.ExpLink(),
method="MLE",
verbose=False): ##{{{
"""
Initialization of the NS Gaussian Model
"""
NSAbstractModel.__init__(self)
self._norm = sd.Normal(method=method)
self._verbose = verbose
self._link = {"loc": link_loc, "scale": link_scale}
self.mu0 = None
self.mu1 = None
self.scale0 = None
self.scale1 = None
self._mut = None
self._scalet = None
def from_netcdf(ifile, ns_law=None): ##{{{
with nc.Dataset(ifile, "r") as ncFile:
## Extract dimensions
##===================
time = np.ma.getdata(ncFile.variables["time"][:])
sample = ncFile.variables["sample"][:]
forcing = ncFile.variables["forcing"][:]
models = ncFile.variables["models"][:]
ns_param = ncFile.variables["ns_param"][:]
stats = ncFile.variables["stat"][:]
reference = np.ma.getdata(ncFile.variables["reference"][:])
if ns_law is None:
## Extract ns_law attributes
##==========================
ns_law_attr = []
for p in ncFile.__dict__:
if "ns_law" in p: ns_law_attr.append(p)
## Transform ns_law attributes to ns_law params
##=============================================
ns_law_kwargs = {}
for p in ns_law_attr:
if "ns_law_param" in p:
pn, pk = p.split("_")[-2:]
if pk == "cst":
ns_law_kwargs[pn +
"_cst"] = ncFile.__dict__[p] == str(True)
if pk == "link":
if ncFile.__dict__[p] == str(sdt.ExpLink()):
ns_law_kwargs["l_" + pn] = sdt.ExpLink()
else:
ns_law_kwargs["l_" + pn] = sdt.IdLink()
ns_law = None
if ncFile.__dict__["ns_law_name"] == "Normal":
ns_law = Normal(**ns_law_kwargs)
elif ncFile.__dict__["ns_law_name"] == "GEV":
ns_law = GEV(**ns_law_kwargs)
elif ncFile.__dict__["ns_law_name"] == "GEVMin":
ns_law = GEVMin(**ns_law_kwargs)
## Set climatology
##=================
clim = Climatology(time, models, ns_law)
clim.X = xr.DataArray(np.ma.getdata(ncFile.variables["X"][:]),
coords=[time, sample, forcing, models],
dims=["time", "sample", "forcing", "models"])
clim.ns_params = xr.DataArray(np.ma.getdata(
ncFile.variables["ns_params"][:]),
coords=[ns_param, sample, models],
dims=["ns_params", "sample", "models"])
clim.stats = xr.DataArray(np.ma.getdata(ncFile.variables["stats"][:]),
coords=[time, sample, stats, models],
dims=["time", "sample", "stats", "models"])
clim.mm_params.mean = np.ma.getdata(ncFile.variables["mm_mean"][:])
clim.mm_params.cov = np.ma.getdata(ncFile.variables["mm_cov"][:])
## Set event
event = Event(ncFile.event_name,
reference.dtype.type(ncFile.event_time),
clim.X.dtype.type(ncFile.event_anomaly), reference,
ncFile.event_type, ncFile.event_side,
ncFile.event_variable, ncFile.event_unit)
return clim, event
if __name__ == "__main__": ## Define dataset ##=============== t,X,_,_ = sdt.Dataset.covariates(2000) loc = 1 - 2 * X scale = 0.5 + 0.5 * X shape = np.repeat( -0.2 , t.size ) Y = - sc.genextreme.rvs( loc = loc , scale = scale , c = -shape ) ## mod = nsm.GEVMin( l_scale = sdt.IdLink() ) mod.fit( Y , X ) mod.set_covariable( X , t ) ub = mod.upper_boundt(t) lb = mod.lower_boundt(t) Y2 = mod.rvs(t) print(mod.get_params()) ## Plot ##===== nrow,ncol = 2,2 fig = plt.figure( figsize = (12,8) ) ax = fig.add_subplot( nrow , ncol , 1 ) ax.plot( t , X , color = "red" ) ax.plot( t , Y , color = "blue" , linestyle = "" , marker = "." )