def get_one_sigma_errs(indep_vars, params, model, **kwargs):
"""
Return the function mean and 1-sigma error at each `k` value
"""
data = {name : params[name] for name in params.columns}
data.update(kwargs)
def function(data, **indep_vars):
data.update(indep_vars)
values = model(**data)
errs = one_sigma(values)
return 0.5*(errs[0] + errs[1])
names = indep_vars.keys()
indep_vars = np.vstack([v for v in indep_vars.values()]).T
return np.array([function(data, **dict(zip(names, row))) for row in indep_vars])
def load_data(data, space, **meta):
"""
Load the data, saved in columns to an ASCII file, into either
a ``PkmuResult`` or ``PkResult`` object
"""
# make the edges
k = data[:, 0]
kedges = make_kedges(data)
factor = 1.0 / 12 ** 0.5
if space == "redshift":
Nmu = 5
muedges = np.linspace(0, 1, Nmu + 1)
mu = 0.5 * (muedges[1:] + muedges[:-1])
k, mu = np.broadcast_arrays(k[..., None], mu[None, ...])
power = np.vstack([data[:, 2 * i + 1] for i in range(Nmu)]).T
error = np.vstack([data[:, 2 * (i + 1)] for i in range(Nmu)]).T
data_dict = {"power": power, "error": error * factor, "k": k, "mu": mu}
return Power2dDataSet([kedges, muedges], data_dict, **meta)
elif space == "real":
data_dict = {"power": data[:, -2], "error": data[:, -1] * factor, "k": k}
return Power1dDataSet(kedges, data_dict, **meta)
else:
raise ValueError
def write(self, key, result, model, eval_kws, **extra):
if 'k' not in key:
raise ValueError("to write the best-fit function must have `k` in the key")
index_cols = key.keys()
with self.open(index_cols) as output:
Nk = len(key['k'])
# the function mean
kws = eval_kws.copy()
params = dict(zip(result.free_names, result.values()))
kws.update(params)
mu = model.eval(**kws)
# get the 1sigma errors
traces = [result[k].flat_trace for k in result.free_names]
other_names = [name for name in model.param_names if name not in result.free_names]
traces += [np.repeat(result[name].median, len(traces[0])) for name in other_names]
params = np.vstack(traces).T
params = pd.DataFrame(params, columns=result.free_names+other_names)
indep_vars = {k:eval_kws[k] for k in model.independent_vars}
mean_errs = get_one_sigma_errs(indep_vars, params, model.eval, **eval_kws)
# append to the output frame
d = {'mean':mu, 'error':mean_errs}
for k in key:
if np.isscalar(key[k]):
d[k] = np.repeat(key[k], Nk)
else:
d[k] = key[k]
for k in extra:
d[k] = np.repeat(extra[k], Nk)
self._output = output.append(pd.DataFrame(d))
