def take_average(self):
anoms_by_grid = {}
for key in self.obs_by_grid:
anoms_by_grid[key] = [qc.winsorised_mean(self.obs_by_grid[key]),
len(self.obs_by_grid[key])]
self.obs_by_grid = {}
return anoms_by_grid
def grid_super_obs(reps, intype):
'''
Aggregate a list of :class:`.MarineReport` onto a 1x1x5-day grid using resistant averages
:param reps: list of MarineReports to be gridded
:param intype: one of either 'SST' for gridding SST or 'MAT' for gridding MAT
:type reps: :class:`.MarineReport`
:type intype: string
:return: a dictionary whose keys are :class:`.Gridpt` instances and which contain two-element lists with the robust average and the number of observations contributing to it.
:rtype: dictionary of list of floats. Keys are :class:`.Gridpt`
The function grids anomalies from a list of marine reports. To save space and generally speed things up,
the whole grid is never calculated. Instead, a dictionary is used. The keys are the populated grid boxes
(as :class:`.Gridpt` objects) and lists are stored for each key. The output is in this format with the list under each key
containing an average anomaly and a count of observations.
'''
assert (intype == 'SST' or intype == 'AT'), intype
# this does a simple gridding of observations of type "MarineReport"
# with observations and averages for each grid cell held in dictionaries
# it returns a dictionary with an entry for each occupied grid cell
# and each entry is a list with two elements:
# the gridcell average and the number of obs contributing to that average
#make dictionary (obs_by_grid) of all anomalies referenced by grid point
obs_by_grid = {}
for this_report in reps:
rep = Gridpt(this_report.getvar('LAT'), this_report.getvar('LON'),
this_report.getvar('YR'), this_report.getvar('MO'),
this_report.getvar('DY'))
x = this_report.getanom(intype)
if x != None:
if rep in obs_by_grid:
obs_by_grid[rep].append(x)
else:
obs_by_grid[rep] = [x]
#calculate the trimmed mean of the super obs in each grid box
#loop over all the keys in obs_by_grid. Each key is the
#index for a different grid box
#and each entry in obs_by_grid is a list of anomalies for
#which we want to take the trimmed mean
#and count the available observations
anoms_by_grid = {}
for key in obs_by_grid:
anoms_by_grid[key] = [qc.winsorised_mean(obs_by_grid[key]),
len(obs_by_grid[key])]
return anoms_by_grid
def test_longer_ascending_run_two_outliers(self):
inarr = [
1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., 12., 13., 14., 99.,
1000.
]
self.assertEqual(8.5, qc.winsorised_mean(inarr))
def test_longer_ascending_run_large_outlier(self):
inarr = [1., 2., 3., 4., 5., 6., 700.]
self.assertEqual(4.0, qc.winsorised_mean(inarr))
def test_ascending(self):
inarr = [3., 4., 5.]
self.assertEqual(4.0, qc.winsorised_mean(inarr))
def test_three_fours(self):
inarr = [4.0, 4.0, 4.0]
self.assertEqual(4.0, qc.winsorised_mean(inarr))
def test_all_zeroes(self):
inarr = [0, 0, 0, 0, 0, 0, 0, 0, 0]
self.assertEqual(0.0, qc.winsorised_mean(inarr))
