def adeltap(self,tsl,N):
_n = int(N)
logger.debug("L'esponente = %d",_n)
# if _n not in (4,12):
# logger.error("L'esponente della funzione ADELTAP può essere solo 12 o 4 invece è %d",_n)
# return tsl
_ts = tsl._data
_t = ts.tshift(_ts,-1,True)
_v = (((_ts/_t)**_n)-1.0)*100.0
_v[_v.mask] = np.nan
tsl._data = _v
tsl._data.mask = None
return tsl
def shift(self,tsl,N):
"""Shift the time-series (lag/lead)
:param tsl: the input time-series
:type tsl: a time-series
:param N: N>0 lag N<0 lead
:type N: integer
"""
_n = int(N)
_ts = tsl._data
_t = ts.tshift(_ts,_n,True)
tsl._data = _t
return tsl
def set_indices(self, full_year=None, reference_season=None, minimum_size=None, lag=0):
"""
Sets the ENSO indices for the ENSO indicator.
Results are cached for convenience: if the method is called without arguments,
the monthly indices are returned, provided they had been already computed.
Parameters
----------
full_year : {None, False, True}, optional
Whether the ENSO indices are set for a period of 12 months (True) or by
valid episodes (False).
If None, defaults to :attr:`full_year`.
reference_season : {None, string or sequence}, optional
Reference season.
If None, defaults to :attr:`reference_season`.
minimum_size : {None, int}, optional
Minimum size for the groups of consecutive values.
If None, defaults to :attr:`minimum_size`.
lag : {integer}, optional
Number of months of lag for ENSO indices. For example,
if lag=6, the ENSO phase starting in Oct. 2001 is applied starting on
Apr. 2002.
Notes
-----
* If ``full_year`` is False, this method calls :meth:`set_monthly_indices`.
Otherwise, it calls :meth:`set_annual_indices`.
* Results are cached for convenience.
The cached results are always with a monthly frequency, and converted to
the current frequency ``freq`` as needed, with the command
>>> scikits.timeseries.lib.backward_fill(_cached.convert(freq, func=ma.mean))
* The cached indices have a monthly frequency by default.
When the ENSOIndicator is converted to a lower frequency with a given
conversion function, the ENSO indices are converted using
:func:`scipy.stats.mstats.mode` as conversion function.
"""
optinfo = self._optinfo
# Check the full_year flag: reset _cachedcurrent if it changed....
if full_year is None:
full_year = optinfo.get("full_year", False)
else:
full_year = bool(full_year)
if optinfo.get("full_year", False) != full_year:
optinfo["full_year"] = full_year
self._cachedcurrent = None
# Check the reference season .....................................
if reference_season is None:
reference_season = optinfo.get("reference_season", None)
else:
optreference_season = optinfo.get("reference_season", ())
if tuple(reference_season) != tuple(optreference_season):
self.refseason = reference_season
# Check the minimum_size..........................................
if minimum_size is None:
minimum_size = optinfo.get("minimum_size", None)
else:
self.minimum_size = minimum_size
# Check the current lag ..........................................
if lag != optinfo.get("current_lag", 0):
optinfo["current_lag"] = lag
self._cachedcurrent = None
# Define what we need to get......................................
if full_year:
_cached = self._cachedmonthly.get("indices_annual", None)
func = self._set_annual_indices
else:
_cached = self._cachedmonthly.get("indices_monthly", None)
func = self._set_monthly_indices
# ...............
if _cached is None:
if (minimum_size is None) and (reference_season is None):
err_msg = (
"The ENSO indices have not been initialized!\n"
"Please define a reference season with `self.refseason`"
" and a minimum cluster size with `self.minsize`"
)
raise ValueError(err_msg)
_cached = func(reference_season=reference_season, minimum_size=minimum_size)
# Make sure we didn't zap full_year
optinfo["full_year"] = full_year
# Check whether we need to take a lag into account
if lag:
_cached = ts.tshift(func(reference_season=reference_season, minimum_size=minimum_size), -lag, copy=False)
# Fix the frequency/date range/shape of _cached
_cached = self._fix_cachedcurrent(_cached)
self._cachedcurrent = _cached
return _cached
def set_indices(self, full_year=None, reference_season=None,
minimum_size=None, lag=0):
"""
Sets the ENSO indices for the ENSO indicator.
Results are cached for convenience: if the method is called without arguments,
the monthly indices are returned, provided they had been already computed.
Parameters
----------
full_year : {None, False, True}, optional
Whether the ENSO indices are set for a period of 12 months (True) or by
valid episodes (False).
If None, defaults to :attr:`full_year`.
reference_season : {None, string or sequence}, optional
Reference season.
If None, defaults to :attr:`reference_season`.
minimum_size : {None, int}, optional
Minimum size for the groups of consecutive values.
If None, defaults to :attr:`minimum_size`.
lag : {integer}, optional
Number of months of lag for ENSO indices. For example,
if lag=6, the ENSO phase starting in Oct. 2001 is applied starting on
Apr. 2002.
Notes
-----
* If ``full_year`` is False, this method calls :meth:`set_monthly_indices`.
Otherwise, it calls :meth:`set_annual_indices`.
* Results are cached for convenience.
The cached results are always with a monthly frequency, and converted to
the current frequency ``freq`` as needed, with the command
>>> scikits.timeseries.lib.backward_fill(_cached.convert(freq, func=ma.mean))
* The cached indices have a monthly frequency by default.
When the ENSOIndicator is converted to a lower frequency with a given
conversion function, the ENSO indices are converted using
:func:`scipy.stats.mstats.mode` as conversion function.
"""
optinfo = self._optinfo
# Check the full_year flag: reset _cachedcurrent if it changed....
if full_year is None:
full_year = optinfo.get('full_year', False)
else:
full_year = bool(full_year)
if optinfo.get('full_year', False) != full_year:
optinfo['full_year'] = full_year
self._cachedcurrent = None
# Check the reference season .....................................
if reference_season is None:
reference_season = optinfo.get('reference_season', None)
else:
optreference_season = optinfo.get('reference_season', ())
if tuple(reference_season) != tuple(optreference_season):
self.refseason = reference_season
# Check the minimum_size..........................................
if minimum_size is None:
minimum_size = optinfo.get("minimum_size", None)
else:
self.minimum_size = minimum_size
# Check the current lag ..........................................
if lag != optinfo.get('current_lag', 0):
optinfo['current_lag'] = lag
self._cachedcurrent = None
# Define what we need to get......................................
if full_year:
_cached = self._cachedmonthly.get('indices_annual', None)
func = self._set_annual_indices
else:
_cached = self._cachedmonthly.get('indices_monthly', None)
func = self._set_monthly_indices
#...............
if _cached is None:
if (minimum_size is None) and (reference_season is None):
err_msg = "The ENSO indices have not been initialized!\n"\
"Please define a reference season with `self.refseason`"\
" and a minimum cluster size with `self.minsize`"
raise ValueError(err_msg)
_cached = func(reference_season=reference_season,
minimum_size=minimum_size)
# Make sure we didn't zap full_year
optinfo['full_year'] = full_year
# Check whether we need to take a lag into account
if lag:
_cached = ts.tshift(func(reference_season=reference_season,
minimum_size=minimum_size),
- lag, copy=False)
# Fix the frequency/date range/shape of _cached
_cached = self._fix_cachedcurrent(_cached)
self._cachedcurrent = _cached
return _cached
def deltap(self,tsl,n=12):
_ts = tsl._data
_n = int(n)
_t = ts.tshift(_ts,-_n,False)
tsl._data = (_ts/_t-1.0)*100.0
return tsl