def load_parameters_from_python_file(filename, as_module=False):
if not as_module:
f = support.check_file_existence(filename)
try:
module = imp.load_source('module', f)
except Exceprint, e:
support.error('File ' + filename + " is not valid python file.",
error_instance=e)
def main():
parser = argparse.ArgumentParser(
'GADMA module for runs of local search on bootstrapped data. Is needed for calculating confidence intervals.'
)
parser.add_argument('-b',
'--boots',
metavar="<dir>",
required=True,
help='Directory where bootstrapped data is located.')
parser.add_argument(
'-d',
'--dem_model',
metavar="<filename>",
required=True,
help=
'File with demographic model. Should contain `model_func` or `generated_model` function. One can put there several extra parameters and they will be taken automatically, otherwise one will need to enter them manually. Such parameters are:\n\t1) p0 (or popt) - initial parameters values\n\t2) lower_bound - list of lower bounds for parameters values\n\t4) upper_bound - list of upper bounds for parameters values\n\t5) par_labels/param_labels - list of string names for parameters 6) pts - pts for dadi (if there is no pts then moments will be run automatically).'
)
parser.add_argument('-o',
'--output',
metavar="<dir>",
required=True,
help='Output directory.')
parser.add_argument('-j',
'--jobs',
metavar="N",
type=int,
default=1,
help='Number of threads for parallel run.')
parser.add_argument(
'--opt',
metavar="log/powell",
type=str,
default='log',
help=
'Local search algorithm, by now it can be:\n\t1) `log` - Inference.optimize_log\n\t2) `powell` - Inference.optimize_powell.'
)
parser.add_argument(
'-p',
'--params',
metavar="<filename>",
type=str,
default=None,
help=
'Filename with parameters, should be valid python file. Parameters are presented in -d/--dem_model option description upper.'
)
args = parser.parse_args()
output = support.ensure_dir_existence(args.output, False)
all_boot = gadma.Inference.load_bootstrap_data_from_dir(
args.boots, return_filenames=True)
print(str(len(all_boot)) + ' bootstrapped data found.')
#extract model_func, we cannot put it to function as multiprocessing need pickleable functions
args.dem_model = support.check_file_existence(args.dem_model)
try:
file_with_model_func = imp.load_source('module', args.dem_model)
except Exception, e:
support.error('File ' + args.dem_model + " is not valid python file.",
error_instance=e)
def main():
parser = argparse.ArgumentParser(
'GADMA module for calculating confidence intervals from the result table of local search runs on bootstrapped data.'
)
parser.add_argument(
'input_filename',
metavar='<filename>',
help=
'Filename (.csv or .pkl) with result from local search runs on bootstrapped data. Output of gadma-run_ls_on_boot_data.'
)
parser.add_argument(
'--log',
required=False,
action='store_true',
help=
'If log then logarithm will be used to calculate confidence intervals.'
)
parser.add_argument('--tex',
required=False,
action='store_true',
help='Tex output.')
parser.add_argument('--acc',
required=False,
metavar='N',
type=int,
default=5,
help='Accuracy of output (dafault: 5).')
args = parser.parse_args()
filename = support.check_file_existence(args.input_filename)
ext = filename.split('.')[-1]
if ext == 'csv':
df = pd.read_csv(filename, index_col=0)
elif ext == 'pkl':
df = pd.read_pickle(filename)
a = df.values
if args.log:
a = np.log(a)
means = np.mean(a, axis=0)
stds = np.std(a, axis=0)
for m, s, x, par_name in zip(means, stds, a.T, df.columns):
l = m - 1.96 * s
u = m + 1.96 * s
if args.log:
l = np.exp(l)
u = np.exp(u)
k2, p = stats.normaltest(x)
if p < 0.05:
normtest_str = 'data looks '
if args.log:
normtest_str += 'log-normal (fail to reject H0)'
else:
normtest_str += 'normal (fail to reject H0)'
else:
normtest_str = 'data does not look '
if args.log:
normtest_str += 'log-normal (reject H0)'
else:
normtest_str += 'normal (reject H0)'
normtest_str += ' p-value=%.2e' % p
if args.tex:
format_string = '%s:\t$[%.' + str(args.acc) + 'f - %.' + str(
args.acc) + 'f]$\t%s'
else:
format_string = '%s:\t%.' + str(args.acc) + 'f\t%.' + str(
args.acc) + 'f\t%s'
print(format_string % (par_name.strip(), l, u, normtest_str))