def gen_new_fake_data(n_rain=50, n_samples=10,
rel_noise=2, mean_slope=5):
dummy_rain = np.array([[abs(x)] for x in gauss(50)])
arr2csv(dummy_rain, 'dummy_rainfalldata.csv')
sample_sizes = np.random.randint(n_rain//4, n_rain*2, n_samples)
if opt.TEST_FALSE_CASE:
dummy_samples = [abs(gauss(sz)) * rel_noise for sz in sample_sizes]
else:
max_shift = opt.MAX_FAKE_DATA_SHIFT
shifts = np.random.randint(-max_shift, max_shift, n_samples)
slopes = np.array([abs(x) for x in gauss(n_samples)]) + mean_slope
constants = [abs(x) for x in gauss(n_samples)]
dummy_samples = []
for idx, shift in enumerate(shifts):
sample_size = sample_sizes[idx]
sample = gauss(sample_size) * rel_noise # add noise
rainyr_first = shift
rainyr_last = shift + n_rain - 1
transformed_rain = slopes[idx] * dummy_rain + constants[idx]
if shift > 0:
time_period = range(shift, shift + n_rain)
for rain_yr, yr in enumerate(time_period):
try:
# sample[shift] <-- rain[0]
sample[yr] += transformed_rain[rain_yr]
except IndexError:
continue
else:
time_period = range(abs(shift), abs(shift) + n_rain)
for yr, rain_yr in enumerate(time_period):
try:
# sample[0] <-- rain[shift]
sample[yr] += transformed_rain[rain_yr]
except IndexError:
continue
dummy_samples.append([abs(x) for x in sample])
arr2csv([shifts, slopes, constants], 'dummy_params.csv')
arr2csv(transpose(dummy_samples), 'dummy_sampledata.csv')
def gen_new_fake_data(n_rain=50, n_samples=10, rel_noise=2, mean_slope=5):
dummy_rain = np.array([[abs(x)] for x in gauss(50)])
arr2csv(dummy_rain, 'dummy_rainfalldata.csv')
sample_sizes = np.random.randint(n_rain // 4, n_rain * 2, n_samples)
if opt.TEST_FALSE_CASE:
dummy_samples = [abs(gauss(sz)) * rel_noise for sz in sample_sizes]
else:
max_shift = opt.MAX_FAKE_DATA_SHIFT
shifts = np.random.randint(-max_shift, max_shift, n_samples)
slopes = np.array([abs(x) for x in gauss(n_samples)]) + mean_slope
constants = [abs(x) for x in gauss(n_samples)]
dummy_samples = []
for idx, shift in enumerate(shifts):
sample_size = sample_sizes[idx]
sample = gauss(sample_size) * rel_noise # add noise
rainyr_first = shift
rainyr_last = shift + n_rain - 1
transformed_rain = slopes[idx] * dummy_rain + constants[idx]
if shift > 0:
time_period = range(shift, shift + n_rain)
for rain_yr, yr in enumerate(time_period):
try:
# sample[shift] <-- rain[0]
sample[yr] += transformed_rain[rain_yr]
except IndexError:
continue
else:
time_period = range(abs(shift), abs(shift) + n_rain)
for yr, rain_yr in enumerate(time_period):
try:
# sample[0] <-- rain[shift]
sample[yr] += transformed_rain[rain_yr]
except IndexError:
continue
dummy_samples.append([abs(x) for x in sample])
arr2csv([shifts, slopes, constants], 'dummy_params.csv')
arr2csv(transpose(dummy_samples), 'dummy_sampledata.csv')
# Data Pre-Processing
############################################################
assert 0 <= opt.MIN_OVERLAP <= len(csv2arr(opt.RAINFALLDATA_FN))
# format and normalize rainfall data
rain = csv2arr(opt.RAINFALLDATA_FN, hasheaders=opt.RAIN_DATA_HAS_HEADERS)
rain = list(np.array(rain).T[0]) # convert Nx1 array to list
rain = [float(x) for x in rain]
# read and format sample data and take the transpose
# (so now each sample is a row)
sampledata = csv2arr(opt.SAMPLEDATA_FN, hasheaders=opt.SAMPLE_DATA_HAS_HEADERS)
# take the transpose (so now each sample is a row)
sampledata = transpose(sampledata)
# convert any strings to floats
sampledata = [[float(x) for x in sample if x] for sample in sampledata]
############################################################
# Analysis
############################################################
if not opt.DEBUG_MODE_ON or opt.TEST_DUMMY_RESULTS_LIKE_THEY_ARE_REAL:
# Test single sample results if using generated fake data
single_sample_results, probs = align(rain, sampledata)
print "\nSingle Sample Results:"
q = 1
print "# : (p < {}, p, r, shift)".format(opt.ALPHA)
digs = int(ceil(log10(len(single_sample_results))))
for idx, res in enumerate(single_sample_results):
# if None, defaults to 'summary_data.csv' (in the folder containing this
# script)
outf = None
#####################################################
# set defaults
if not rows2grab:
rows2grab = [2, 3]
if not trans2grab:
trans2grab = range(10)
if not summary_dir:
summary_dir = os.path.join(os.getcwd(), 'summaries')
if not outf:
outf = 'summary_data.csv'
# do stuff
data = []
data_guide = []
for fn in os.listdir(summary_dir):
path2summary = os.path.join(summary_dir, fn)
summary = csv2arr(path2summary)[4:] # throw out first 4 rows
for m in trans2grab:
for k in rows2grab:
data.append(summary[11*m + k][3:])
x = 10*m + k + 4 # in original csv (not counting blank rows)
data_guide.append(fn[:-4] + '_({}-{}-{}))'.format(m, k, x))
data = transpose(data)
data.insert(0, data_guide)
arr2csv(data, filename=outf)
# if None, defaults to 'summary_data.csv' (in the folder containing this
# script)
outf = None
#####################################################
# set defaults
if not rows2grab:
rows2grab = [2, 3]
if not trans2grab:
trans2grab = range(10)
if not summary_dir:
summary_dir = os.path.join(os.getcwd(), 'summaries')
if not outf:
outf = 'summary_data.csv'
# do stuff
data = []
data_guide = []
for fn in os.listdir(summary_dir):
path2summary = os.path.join(summary_dir, fn)
summary = csv2arr(path2summary)[4:] # throw out first 4 rows
for m in trans2grab:
for k in rows2grab:
data.append(summary[11 * m + k][3:])
x = 10 * m + k + 4 # in original csv (not counting blank rows)
data_guide.append(fn[:-4] + '_({}-{}-{}))'.format(m, k, x))
data = transpose(data)
data.insert(0, data_guide)
arr2csv(data, filename=outf)
############################################################
assert 0 <= opt.MIN_OVERLAP <= len(csv2arr(opt.RAINFALLDATA_FN))
# format and normalize rainfall data
rain = csv2arr(opt.RAINFALLDATA_FN, hasheaders=opt.RAIN_DATA_HAS_HEADERS)
rain = list(np.array(rain).T[0]) # convert Nx1 array to list
rain = [float(x) for x in rain]
# read and format sample data and take the transpose
# (so now each sample is a row)
sampledata = csv2arr(opt.SAMPLEDATA_FN, hasheaders=opt.SAMPLE_DATA_HAS_HEADERS)
# take the transpose (so now each sample is a row)
sampledata = transpose(sampledata)
# convert any strings to floats
sampledata = [[float(x) for x in sample if x] for sample in sampledata]
############################################################
# Analysis
############################################################
if not opt.DEBUG_MODE_ON or opt.TEST_DUMMY_RESULTS_LIKE_THEY_ARE_REAL:
# Test single sample results if using generated fake data
single_sample_results, probs = align(rain, sampledata)
print "\nSingle Sample Results:"
q = 1
print "# : (p < {}, p, r, shift)".format(opt.ALPHA)
digs = int(ceil(log10(len(single_sample_results))))
for idx, res in enumerate(single_sample_results):