def __errors_repl(subset=None, verbosity=0):
"""Calculate the errors for peak intensities from replicated spectra.
@keyword subset: The list of spectrum ID strings to restrict the analysis to.
@type subset: list of str
@keyword verbosity: The amount of information to print. The higher the value, the greater the verbosity.
@type verbosity: int
"""
# Replicated spectra.
repl = replicated_flags()
# Are all spectra replicated?
if False in list(repl.values()):
all_repl = False
print("All spectra replicated: No.")
else:
all_repl = True
print("All spectra replicated: Yes.")
# Initialise.
if not hasattr(cdp, 'sigma_I'):
cdp.sigma_I = {}
if not hasattr(cdp, 'var_I'):
cdp.var_I = {}
# The subset.
subset_flag = False
if not subset:
subset_flag = True
subset = cdp.spectrum_ids
# Loop over the spectra.
for id in subset:
# Skip non-replicated spectra.
if not repl[id]:
continue
# Skip replicated spectra which already have been used.
if id in cdp.var_I and cdp.var_I[id] != 0.0:
continue
# The replicated spectra.
for j in range(len(cdp.replicates)):
if id in cdp.replicates[j]:
spectra = cdp.replicates[j]
# Number of spectra.
num_spectra = len(spectra)
# Printout.
print("\nReplicated spectra: " + repr(spectra))
if verbosity:
print("%-20s%-20s" % ("Spin_ID", "SD"))
# Calculate the mean value.
count = 0
for spin, spin_id in spin_loop(return_id=True):
# Skip deselected spins.
if not spin.select:
continue
# Skip and deselect spins which have no data.
if not hasattr(spin, 'peak_intensity'):
spin.select = False
continue
# Missing data.
missing = False
for j in range(num_spectra):
if not spectra[j] in spin.peak_intensity:
missing = True
if missing:
continue
# The peak intensities.
values = []
for j in range(num_spectra):
values.append(spin.peak_intensity[spectra[j]])
# The standard deviation.
sd = std(values=values, dof=1)
# Printout.
if verbosity:
print("%-20s%-20s" % (spin_id, sd))
# Sum of variances (for average).
if not id in cdp.var_I:
cdp.var_I[id] = 0.0
cdp.var_I[id] = cdp.var_I[id] + sd**2
count = count + 1
# No data catch.
if not count:
raise RelaxError(
"No data is present, unable to calculate errors from replicated spectra."
)
# Average variance.
cdp.var_I[id] = cdp.var_I[id] / float(count)
# Set all spectra variances.
for j in range(num_spectra):
cdp.var_I[spectra[j]] = cdp.var_I[id]
# Print out.
print("Standard deviation: %s" % sqrt(cdp.var_I[id]))
# Average across all spectra if there are time points with a single spectrum.
if not all_repl:
# Print out.
if subset_flag:
print("\nVariance averaging over the spectra subset.")
else:
print("\nVariance averaging over all spectra.")
# Initialise.
var_I = 0.0
num_dups = 0
# Loop over all time points.
for id in cdp.var_I:
# Only use id's defined in subset
if id not in subset:
continue
# Single spectrum (or extraordinarily accurate NMR spectra!).
if cdp.var_I[id] == 0.0:
continue
# Sum and count.
var_I = var_I + cdp.var_I[id]
num_dups = num_dups + 1
# Average value.
var_I = var_I / float(num_dups)
# Assign the average value to all time points.
for id in subset:
cdp.var_I[id] = var_I
# Print out.
print("Standard deviation for all spins: " + repr(sqrt(var_I)))
# Loop over the spectra.
for id in cdp.var_I:
# Create the standard deviation data structure.
cdp.sigma_I[id] = sqrt(cdp.var_I[id])
# Set the spin specific errors.
for spin in spin_loop():
# Skip deselected spins.
if not spin.select:
continue
# Set the error.
spin.peak_intensity_err = cdp.sigma_I
def monte_carlo_error_analysis():
"""Function for calculating errors from the Monte Carlo simulations.
The standard deviation formula used to calculate the errors is the square root of the
bias-corrected variance, given by the formula::
__________________________
/ 1
sd = / ----- * sum({Xi - Xav}^2)
\/ n - 1
where
- n is the total number of simulations.
- Xi is the parameter value for simulation i.
- Xav is the mean parameter value for all simulations.
"""
# Test if the current data pipe exists.
check_pipe()
# Test if simulations have been set up.
if not hasattr(cdp, 'sim_state'):
raise RelaxError("Monte Carlo simulations have not been set up.")
# The specific analysis API object.
api = return_api()
# Loop over the models.
for model_info in api.model_loop():
# Get the selected simulation array.
select_sim = api.sim_return_selected(model_info=model_info)
# Loop over the parameters.
index = 0
while True:
# Get the array of simulation parameters for the index.
param_array = api.sim_return_param(index, model_info=model_info)
# Break (no more parameters).
if param_array == None:
break
# Handle dictionary type parameters.
if isinstance(param_array[0], dict):
# Initialise the standard deviation structure as a dictionary.
sd = {}
# Loop over each key.
for key in param_array[0]:
# Create a list of the values for the current key.
data = []
for i in range(len(param_array)):
data.append(param_array[i][key])
# Calculate and store the SD.
sd[key] = statistics.std(values=data, skip=select_sim)
# Handle list type parameters.
elif isinstance(param_array[0], list):
# Initialise the standard deviation structure as a list.
sd = []
# Loop over each element.
for j in range(len(param_array[0])):
# Create a list of the values for the current key.
data = []
for i in range(len(param_array)):
data.append(param_array[i][j])
# Calculate and store the SD.
sd.append(statistics.std(values=data, skip=select_sim))
# SD of simulation parameters with values (ie not None).
elif param_array[0] != None:
sd = statistics.std(values=param_array, skip=select_sim)
# Simulation parameters with the value None.
else:
sd = None
# Set the parameter error.
api.set_error(index, sd, model_info=model_info)
# Increment the parameter index.
index = index + 1
# Turn off the Monte Carlo simulation state, as the MC analysis is now finished.
cdp.sim_state = False
def __errors_repl(subset=None, verbosity=0):
"""Calculate the errors for peak intensities from replicated spectra.
@keyword subset: The list of spectrum ID strings to restrict the analysis to.
@type subset: list of str
@keyword verbosity: The amount of information to print. The higher the value, the greater the verbosity.
@type verbosity: int
"""
# Replicated spectra.
repl = replicated_flags()
# Are all spectra replicated?
if False in list(repl.values()):
all_repl = False
print("All spectra replicated: No.")
else:
all_repl = True
print("All spectra replicated: Yes.")
# Initialise.
if not hasattr(cdp, 'sigma_I'):
cdp.sigma_I = {}
if not hasattr(cdp, 'var_I'):
cdp.var_I = {}
# The subset.
subset_flag = False
if not subset:
subset_flag = True
subset = cdp.spectrum_ids
# Loop over the spectra.
for id in subset:
# Skip non-replicated spectra.
if not repl[id]:
continue
# Skip replicated spectra which already have been used.
if id in cdp.var_I and cdp.var_I[id] != 0.0:
continue
# The replicated spectra.
for j in range(len(cdp.replicates)):
if id in cdp.replicates[j]:
spectra = cdp.replicates[j]
# Number of spectra.
num_spectra = len(spectra)
# Printout.
print("\nReplicated spectra: " + repr(spectra))
if verbosity:
print("%-20s%-20s" % ("Spin_ID", "SD"))
# Calculate the mean value.
count = 0
for spin, spin_id in spin_loop(return_id=True):
# Skip deselected spins.
if not spin.select:
continue
# Skip and deselect spins which have no data.
if not hasattr(spin, 'peak_intensity'):
spin.select = False
continue
# Missing data.
missing = False
for j in range(num_spectra):
if not spectra[j] in spin.peak_intensity:
missing = True
if missing:
continue
# The peak intensities.
values = []
for j in range(num_spectra):
values.append(spin.peak_intensity[spectra[j]])
# The standard deviation.
sd = std(values=values, dof=1)
# Printout.
if verbosity:
print("%-20s%-20s" % (spin_id, sd))
# Sum of variances (for average).
if not id in cdp.var_I:
cdp.var_I[id] = 0.0
cdp.var_I[id] = cdp.var_I[id] + sd**2
count = count + 1
# No data catch.
if not count:
raise RelaxError("No data is present, unable to calculate errors from replicated spectra.")
# Average variance.
cdp.var_I[id] = cdp.var_I[id] / float(count)
# Set all spectra variances.
for j in range(num_spectra):
cdp.var_I[spectra[j]] = cdp.var_I[id]
# Print out.
print("Standard deviation: %s" % sqrt(cdp.var_I[id]))
# Average across all spectra if there are time points with a single spectrum.
if not all_repl:
# Print out.
if subset_flag:
print("\nVariance averaging over the spectra subset.")
else:
print("\nVariance averaging over all spectra.")
# Initialise.
var_I = 0.0
num_dups = 0
# Loop over all time points.
for id in cdp.var_I:
# Only use id's defined in subset
if id not in subset:
continue
# Single spectrum (or extraordinarily accurate NMR spectra!).
if cdp.var_I[id] == 0.0:
continue
# Sum and count.
var_I = var_I + cdp.var_I[id]
num_dups = num_dups + 1
# Average value.
var_I = var_I / float(num_dups)
# Assign the average value to all time points.
for id in subset:
cdp.var_I[id] = var_I
# Print out.
print("Standard deviation for all spins: " + repr(sqrt(var_I)))
# Loop over the spectra.
for id in cdp.var_I:
# Create the standard deviation data structure.
cdp.sigma_I[id] = sqrt(cdp.var_I[id])
# Set the spin specific errors.
for spin in spin_loop():
# Skip deselected spins.
if not spin.select:
continue
# Set the error.
spin.peak_intensity_err = cdp.sigma_I
# Calculate R2eff and store it.
minimise.calculate()
r2eff_indiv[j, i] = spin.r2eff['800.0_%.1f' % spin_lock[j]]
# Randomise I0 once.
int['ref'] = gauss(i0, ERR)
# Calculate all R2eff and store them.
minimise.calculate()
for j in range(len(spin_lock)):
r2eff_group[j, i] = spin.r2eff['800.0_%.1f' % spin_lock[j]]
# The errors.
for j in range(len(spin_lock)):
sigma_r2eff_indiv[j] = std(r2eff_indiv[j])
sigma_r2eff_group[j] = std(r2eff_group[j])
# Plot the data.
file = open('error_plot.agr', 'w')
# Header.
file.write("@with g0\n")
file.write("@ s0 legend \"Full error formula\"\n")
file.write("@ s1 legend \"Reduced error formula\"\n")
file.write("@ s2 legend \"Bootstrap individual dispersion points\"\n")
file.write("@ s3 legend \"Bootstrap group\"\n")
# The full error formula.
file.write("@target G0.S0\n@type xy\n")
for i in range(len(spin_lock)):