def apply_filters(data):
"""
Applies all 4 filters asked for and returns them in a python list.
The list will be structured with rows in the following order:
1. Linear OLS trend
2. HP filters
3. Bandpass filter
4. First-difference
Parameters
----------
data: array-like, dtype=float
The data set that will be run through the filters
Returns
-------
filtered: list
This is a 4 element list with the filtered data in the order
listed above.
Notes
-----
Because of how the filters work, the bandpass and first diff
arrays will be different sizes than the hp filter ones.
Note also that before applying the filters, we take the natural
log of the data.
"""
data = np.log(data)
linear_ols = filters.hp_filter(data, lam=1e10)[1]
hp_fils = filters.hp_filter(data, lam=1600)[1]
bp_fils = filters.bandpass_filter(data, 10, 6, 32)
temp1 = data.copy()
temp1 = np.insert(temp1, temp1.size, 0)
temp2 = data.copy()
temp2 = np.insert(temp2, 0, 0)
yhat = temp1 - temp2
first_diff = yhat[1:-1]
all_fils = [linear_ols, hp_fils, bp_fils, first_diff]
return all_fils
def test_hpfilter(self):
from dolo.numeric.filters import hp_filter
import numpy
import numpy.random
sig = numpy.eye(2) # number of series
N = 100 # number of observations
dumb_series = numpy.random.multivariate_normal([0, 0], sig, N).T
trend_test = hpfilter_dense(dumb_series)
[T, cycle] = hp_filter(dumb_series)
from numpy.testing import assert_almost_equal
assert_almost_equal(trend_test, T)
def test_hpfilter(self):
from dolo.numeric.filters import hp_filter
import numpy
import numpy.random
sig = numpy.eye( 2 ) # number of series
N = 100 # number of observations
dumb_series = numpy.random.multivariate_normal([0,0], sig, N).T
trend_test = hpfilter_dense(dumb_series)
[T, cycle] = hp_filter(dumb_series)
from numpy.testing import assert_almost_equal
assert_almost_equal(trend_test, T)