def monthly_trial(month = 'Dec', phase = 'elnino', n = 30):
from pandas import read_csv
alldivDF, sst, mei, phaseind = monthly_setup(month = month, \
phase = phase, n = n)
fp = base_fp + month + phase + '.txt'
f = open(fp, 'w'); firstline = 'div\trho\tpost\n'; f.write(firstline)
for div in alldivDF:
print div
nipa = NIPAphase(alldivDF[div], sst, mei, phaseind)
nipa.bootcorr(ntim = 50, corrconf = 0.95, quick = False, bootconf = 0.9)
nipa.gridCheck(lim = 5)
nipa.crossvalpcr()
if not nipa.flags['noSST']:
line = '%s\t%.2f\t%i\n' % (div, nipa.correlation, nipa.n_post_grid)
f.write(line)
f.close()
data = read_csv(fp, sep = '\t', index_col = 0)
fig, axes = plt.subplots(2, figsize = (14,14))
for ax, dat in zip(axes.flatten(), data.columns):
fig, ax = valueMap(data[dat], fig = fig, ax = ax)
fig.suptitle(phase + month)
fig.savefig(fp[:-4])
return
def seasonal_trial(season = 'DJF', phase = 'allyears', n = 120):
from pandas import read_csv
alldivDF, sst, mei, phaseind, slp = seasonal_setup(season = season, \
phase = phase, n = n)
print slp.data
fp = base_fp + season + phase + '2.txt'
f = open(fp, 'w'); firstline = 'div\trho\tpost\n'; f.write(firstline)
for div in alldivDF:
print div
nipa = NIPAphase(alldivDF[div], slp, mei, phaseind)
nipa.bootcorr(ntim = 100, corrconf = 0.95, quick = False, bootconf = 0.95)
nipa.gridCheck(lim = 6, debug = True)
nipa.crossvalpcr()
if not nipa.flags['noSST']:
line = '%s\t%.2f\t%i\n' % (div, nipa.correlation, nipa.n_post_grid)
f.write(line)
f.close()
data = read_csv(fp, sep = '\t', index_col = 0)
fig, axes = plt.subplots(2, figsize = (14,14))
for ax, dat in zip(axes.flatten(), data.columns):
fig, ax = valueMap(data[dat], fig = fig, ax = ax)
fig.suptitle(phase + season)
fig.savefig(fp[:-4])
return
def one():
for season in ['DJF', 'MAM']:
kwgroups = create_kwgroups( climdiv_months = months[season], \
sst_lag = sst_lag, n_mon_sst = n_mon_sst, \
mei_lag = mei_lag, n_mon_mei = n_mon_mei, \
slp_lag = slp_lag, n_mon_slp = n_mon_slp, \
climdiv_startyr = startyr, n_yrs = n_yrs, \
)
alldivDF, sst, mei, phaseind, regionalDF, stateDF = get_data(kwgroups)
for n in np.arange(30,75,5):
corrs = {}
pre = {}
post ={}
phase = 'neutral'
phaseind = gen_phase_index(mei, n_yrs = n, phase = phase)
st = time()
for corrconf in [0.95]:
for bootconf in [0.95]:
for div in alldivDF:
print div
clim_data = alldivDF[div]
nipa = NIPAphase(clim_data, sst, mei, phaseind)
nipa.bootcorr(ntim = 100, corrconf = corrconf, bootconf = bootconf)
nipa.gridCheck(lim = 6, debug = False); nipa.gridCheck(lim = 6, debug = False)
nipa.crossvalpcr()
if not nipa.flags['noSST']:
corrs[div] = nipa.correlation
pre[div] = nipa.n_pre_grid
post[div] = nipa.n_post_grid
et = time()
print '%.2f seconds elapsed' % (et-st)
fig, axes = plt.subplots(2, figsize = (14,14))
for ax, dat in zip(axes.flatten(), [corrs, post]):
data = pd.DataFrame.from_dict(dat, orient = 'index')
fig, ax = valueMap(data, fig = fig, ax = ax)
fig.suptitle(phase + season)
if bootconf == 0.9: bc = 90
if bootconf == 0.95: bc = 95
if corrconf == 0.9: cc = 90
if corrconf == 0.95: cc = 95
fp = '/Users/bz/Desktop/temp/%s_%s_%i_%i_%i' % (season, phase, n, cc, bc)
fig.savefig(fp)
plt.close(fig)
return
clim_data = alldivDF[div] nipa = NIPAphase(clim_data, sst, mei, phaseind) nipa.bootcorr(ntim = 200, corrconf = corrconf, bootconf = bootconf) nipa.gridCheck(lim = 6, debug = False); nipa.gridCheck(lim = 6, debug = False) nipa.crossvalpcr() if not nipa.flags['noSST']: corrs[div] = nipa.correlation pre[div] = nipa.n_pre_grid post[div] = nipa.n_post_grid et = time() print '%.2f seconds elapsed' % (et-st) fig, axes = plt.subplots(2, figsize = (14,14)) for ax, dat in zip(axes.flatten(), [corrs, post]): data = pd.DataFrame.from_dict(dat, orient = 'index') fig, ax = valueMap(data, fig = fig, ax = ax) fig.suptitle(phase + season) if bootconf == 0.9: bc = 90 if bootconf == 0.95: bc = 95 if corrconf == 0.9: cc = 90 if corrconf == 0.95: cc = 95 fp = '/Users/bz/Desktop/pshaw/%s_%s_%i_%i_%i' % (season, phase, n, cc, bc) fig.savefig(fp) plt.close(fig) # for i in range(1000):
