def __init__(self, model_file=None, model_dict=None):
assert (model_file is not None) or (model_dict is not None), (
"You have to provide either a model file or a"
"model dictionary")
if model_file is not None:
# Read model file and deserialize into a dictionary
try:
with open(model_file) as f:
self._model_dict = my_yaml.load(f,
Loader=my_yaml.FullLoader)
except IOError:
raise ModelIOError(
"File %s cannot be read. Check path and permissions for current user."
% model_file)
except my_yaml.YAMLError:
raise ModelYAMLError(
"Could not parse file %s. Check your syntax." % model_file)
else:
self._model_dict = model_dict
self._parse()
def _load_one_results(fits_extension):
# Gather analysis type
analysis_type = fits_extension.header.get("RESUTYPE")
# Gather the optimized model
serialized_model = _escape_back_yaml_from_fits(
fits_extension.header.get("MODEL"))
model_dict = my_yaml.load(serialized_model)
optimized_model = ModelParser(model_dict=model_dict).get_model()
# Gather statistics values
statistic_values = collections.OrderedDict()
measure_values = collections.OrderedDict()
for key in fits_extension.header.keys():
if key.find("STAT") == 0:
# Found a keyword with a statistic for a plugin
# Gather info about it
id = int(key.replace("STAT", ""))
value = float(fits_extension.header.get(key))
name = fits_extension.header.get("PN%i" % id)
statistic_values[name] = value
if key.find("MEAS") == 0:
# Found a keyword with a statistic for a plugin
# Gather info about it
id = int(key.replace("MEAS", ""))
name = fits_extension.header.get(key)
value = float(fits_extension.header.get("MV%i" % id))
measure_values[name] = value
if analysis_type == "MLE":
# Get covariance matrix
covariance_matrix = np.atleast_2d(
fits_extension.data.field("COVARIANCE").T)
# Instance and return
return MLEResults(optimized_model,
covariance_matrix,
statistic_values,
statistical_measures=measure_values)
elif analysis_type == "Bayesian":
# Gather samples
samples = fits_extension.data.field("SAMPLES")
# Instance and return
return BayesianResults(optimized_model,
samples.T,
statistic_values,
statistical_measures=measure_values)