def fit_single_parameter(modelid, fullparms, parmid, parmval, noise=False):
"""
Use the full parameter set `fullparms` and leave a single parameter
`parmid` variable during the fit.
Returns the fitted value of the parameter with index `parmid`
"""
corr = Correlation(fit_model=modelid, fit_algorithm=FITALG, verbose=0)
tau = np.exp(np.linspace(np.log(TAUMIN),np.log(TAUMAX), TAULEN))
# Create artificial data by using the current fit_model
data = corr.fit_model(fullparms, tau)
if noise:
if noise is True:
deltanoise = (np.max(data)-np.min(data))/20
else:
deltanoise = (np.max(data)-np.min(data))*noise
anoise = (np.random.random(data.shape[0])-.5)*deltanoise
data += anoise
# Add artificial data to data set
corr.correlation = np.dstack((tau, data))[0]
# Set variable parameters
fit_bool = np.zeros(fullparms.shape[0])
fit_bool[parmid] = True
corr.fit_parameters_variable = fit_bool
fullparms_edit = fullparms.copy()
fullparms_edit[parmid] = parmval
corr.fit_parameters = fullparms_edit
Fit(corr)
return corr.fit_parameters[parmid]
def create_corr():
corr = Correlation()
tau = np.exp(np.linspace(np.log(1e-3),np.log(1e6), 10))
data = corr.fit_model(corr.fit_parameters, tau)
noise = (np.random.random(data.shape[0])-.5)*.0005
data += noise
corr.correlation = np.dstack((tau, data))[0]
return corr
