def simulate(pars, save = False, logs = False):
"""
Simulate data from the colony study using a set of VarroaPop parameters
:param pars: Dictionary of parameters to vary.
ICQueenStrength_mean
ICQueenStrength_sd
ICForagerLifespan_mean
ICForagerLifespan_sd
AIAdultLD50
AIAdultSlope
AILarvaLD50
AILarvaSlope
:return a dictionary of summary stats
"""
parameters = pars.copy() #copy our inputs so that we don't ever modify them (pyabc needs these)
static_pars = {'SimEnd': END_DATE, 'IPollenTrips': 8, 'INectarTrips': 17, 'RQEnableReQueen': 'false'}
for name, value in parameters.items():
if not name.endswith(('_mean','_sd')):
static_pars[name] = value
static_pars['NecPolFileEnable'] = 'true'
weather_path = os.path.join(DATA_DIR,'external/weather/weather_2015','18815_grid_39.875_lat.wea')
all_responses = pd.DataFrame()
for index, site in enumerate(SITES):
exposure_filename = 'neonic_profile_' + site + '.csv'
exposure_path = os.path.join(DATA_DIR,"processed/neonic_profiles/", exposure_filename)#os.path.abspath(os.path.join('data', exposure_filename))
site_pars = generate_start(static_pars.copy(), site, INITIAL_DF)
site_pars['NecPolFileName'] = exposure_path
site_pars['ICQueenStrength'] = 0
site_pars['ICForagerLifespan'] = 0
while not (1 <=site_pars['ICQueenStrength'] <= 5):
site_pars['ICQueenStrength'] = np.random.normal(loc = float(parameters['ICQueenStrength_mean']), scale = float(parameters['ICQueenStrength_sd']))
while not (4 <= site_pars['ICForagerLifespan'] <= 16):
site_pars['ICForagerLifespan'] = np.random.normal(loc = float(parameters['ICForagerLifespan_mean']), scale = float(parameters['ICForagerLifespan_sd']))
vp = VarroaPop(parameters = site_pars, weather_file = weather_path, vrp_file = VRP_FILE,
verbose=False, unique=True, keep_files=save, logs=logs)
vp.run_model()
dates = DATES_STR.iloc[index,:]
site_response = filter_site_responses(vp.get_output(), dates_str= dates)
all_responses = all_responses.append(site_response, ignore_index=True)
#Generate labels for rows and columns
rows = ['_'.join(x) for x in product(SITES, DATES)]
all_responses['Index'] = pd.Series(rows) #Add our row labels
all_responses.set_index("Index", inplace=True) #Set row labels to be the index
filtered_resp = all_responses.loc[:,RESPONSE_FILTER] #Keep only the summary stats that we are using
output_dict = {}
for row in filtered_resp.index:
for col in filtered_resp.columns:
label = "_".join([row,col])
output_dict[label] = filtered_resp.loc[row,col]
return output_dict
def post(self):
clean_files(os.path.abspath('VarroaPy/VarroaPy/files/input'), 7)
clean_files(os.path.abspath('VarroaPy/VarroaPy/files/logs'), 7)
clean_files(os.path.abspath('VarroaPy/VarroaPy/files/output'), 7)
args = parser.parse_args()
params = args['parameters']
params = json.loads(params.replace("'", '"'))
weather = args['weather_file']
vp = VarroaPop(parameters=params,
weather_file=weather,
logs=True,
keep_files=True)
vp.run_model()
output = vp.get_output(json_str=True)
jobID = vp.get_jobID()
return json.loads(output), {"session-id": jobID}
def simulate(pars, save=False, logs=False):
"""
Simulate data from the colony study using a set of VarroaPop parameters
:param pars: Dictionary of parameters to vary.
ICQueenStrength_mean
ICQueenStrength_sd
ICForagerLifespan_mean
ICForagerLifespan_sd
AIAdultLD50
AIAdultSlope
AILarvaLD50
AILarvaSlope
:return a dictionary of summary stats
"""
#print(DATA_DIR)
#print(INITIAL_DF)
#print(parameters)
parameters = pars.copy(
) #copy our inputs so that we don't ever modify them (pyabc needs these)
static_pars = {
'SimStart': START_DATE,
'SimEnd': END_DATE,
'IPollenTrips': 8,
'INectarTrips': 17,
'AIHalfLife': 25,
'RQEnableReQueen': 'false'
}
for name, value in parameters.items():
if not name.endswith(('_mean', '_sd')):
static_pars[name] = value
static_pars['NecPolFileEnable'] = 'true'
weather_path = os.path.join(DATA_DIR, 'weather/15055_grid_35.875_lat.wea')
#all_responses = pd.DataFrame(index = rows, columns = cols)
all_responses = pd.DataFrame()
for index, trt in enumerate(TREATMENTS):
trt_responses_mean = np.empty((len(DATES), len(RESPONSE_VARS)))
trt_responses_sd = np.empty((len(DATES), len(RESPONSE_VARS)))
trt_pars = static_pars.copy()
exposure_filename = 'clo_feeding_' + trt + '.csv'
exposure_path = os.path.join(DATA_DIR, 'food_concentrations',
exposure_filename)
trt_pars['NecPolFileName'] = exposure_path
reps = REPS[index]
rep_responses = np.empty(
([len(DATES), len(RESPONSE_VARS), reps
])) # survey dates (rows) x output vars (cols) x reps (z axis)
for rep in range(0, reps):
vp_pars = generate_start(trt_pars.copy(), trt)
#generate random gaussian parameters
vp_pars['ICQueenStrength'] = 0
vp_pars['ICForagerLifespan'] = 0
while not (1 <= vp_pars['ICQueenStrength'] <= 5):
vp_pars['ICQueenStrength'] = np.random.normal(
loc=float(parameters['ICQueenStrength_mean']),
scale=float(parameters['ICQueenStrength_sd']))
while not (4 <= vp_pars['ICForagerLifespan'] <= 16):
vp_pars['ICForagerLifespan'] = np.random.normal(
loc=float(parameters['ICForagerLifespan_mean']),
scale=float(parameters['ICForagerLifespan_sd']))
vp = VarroaPop(parameters=vp_pars,
weather_file=weather_path,
vrp_file=VRP_FILE,
verbose=False,
unique=True,
keep_files=save,
logs=logs)
vp.run_model()
if trt == "160":
dates = DATES_STR_HIGH
else:
dates = DATES_STR
rep_responses[:, :, rep] = filter_rep_responses(vp.get_output(),
dates_str=dates)
mean_cols = [var[0] + "_mean" for var in RESPONSE_VARS]
sd_cols = [var[0] + "_sd" for var in RESPONSE_VARS]
trt_responses_mean = pd.DataFrame(np.mean(rep_responses, axis=2),
columns=mean_cols)
trt_responses_sd = pd.DataFrame(
np.std(rep_responses, axis=2, ddof=1),
columns=sd_cols) # Note: uses sample sd, not population sd
trt_responses = pd.concat([trt_responses_mean, trt_responses_sd],
axis=1)
#print("Treatment {} responses: {}".format(trt,trt_responses))
start_row = len(DATES) * index
end_row = start_row + len(DATES)
#all_responses.loc[start_row:end_row,:] = trt_responses
all_responses = all_responses.append(trt_responses, ignore_index=True)
#Generate labels for rows and columns
rows = ['_'.join(x) for x in product(TREATMENTS, DATES)]
#print('Row labels: {}'.format(rows))
response_cols = [x[0] for x in RESPONSE_VARS]
cols = ['_'.join([x, y]) for y in ['mean', 'sd'] for x in response_cols]
#print('Col labels: {}'.format(cols))
all_responses['Index'] = pd.Series(rows) # Add our row labels
all_responses.set_index("Index",
inplace=True) # Set row labels to be the index
#print('Final result: {}'.format(all_responses))
filtered_resp = all_responses.loc[:,
RESPONSE_FILTER] # Keep only the summary stats that we are using
#print('Filtered result: {}'.format(filtered_resp))
output_dict = {}
for row in filtered_resp.index:
for col in filtered_resp.columns:
label = "_".join([row, col])
output_dict[label] = filtered_resp.loc[row, col]
#print('Output dictionary: {}'.format(output_dict))
return output_dict
def simulate_all_dates(pars, save=False, logs=False):
"""
Simulate data from the colony study using a set of VarroaPop parameters.
This version returns the population at every single date. This will be
used to create posterior predictive plots.
:param pars: Dictionary of parameters to vary.
ICQueenStrength_mean
ICQueenStrength_sd
ICForagerLifespan_mean
ICForagerLifespan_sd
AIAdultLD50
AIAdultSlope
AILarvaLD50
AILarvaSlope
:return a dict of arrays of varroapop summary stats(treatment x day) for adult/pupae/larvae/eggs/all/pollen/nectar)
"""
start = datetime.datetime.strptime(START_DATE, "%m/%d/%Y")
end = datetime.datetime.strptime(DATES_STR_HIGH[3], "%m/%d/%Y")
all_dates = [(start + datetime.timedelta(days=x)).strftime("%m/%d/%Y")
for x in range(0, (end - start).days)]
parameters = pars.copy(
) # copy our inputs so that we don't ever modify them (pyabc needs these)
static_pars = {
'SimStart': START_DATE,
'SimEnd': END_DATE,
'IPollenTrips': 8,
'INectarTrips': 17,
'AIHalfLife': 25,
'RQEnableReQueen': 'false'
}
for name, value in parameters.items():
if not name.endswith(('_mean', '_sd')):
static_pars[name] = value
static_pars['NecPolFileEnable'] = 'true'
weather_path = os.path.join(DATA_DIR, 'weather/15055_grid_35.875_lat.wea')
#all_responses = pd.DataFrame(index = rows, columns = cols)
all_responses = pd.DataFrame()
for index, trt in enumerate(TREATMENTS):
trt_responses_mean = np.empty((len(DATES), len(RESPONSE_VARS)))
trt_responses_sd = np.empty((len(DATES), len(RESPONSE_VARS)))
trt_pars = static_pars.copy()
exposure_filename = 'clo_feeding_' + trt + '.csv'
exposure_path = os.path.join(DATA_DIR, 'food_concentrations',
exposure_filename)
trt_pars['NecPolFileName'] = exposure_path
reps = REPS[index]
rep_responses = np.empty(
([len(all_dates), len(RESPONSE_VARS), reps
])) # survey dates (rows) x output vars (cols) x reps (z axis)
for rep in range(0, reps):
vp_pars = generate_start(trt_pars.copy(), trt)
#generate random gaussian parameters
vp_pars['ICQueenStrength'] = 0
vp_pars['ICForagerLifespan'] = 0
while not (1 <= vp_pars['ICQueenStrength'] <= 5):
vp_pars['ICQueenStrength'] = np.random.normal(
loc=float(parameters['ICQueenStrength_mean']),
scale=float(parameters['ICQueenStrength_sd']))
while not (4 <= vp_pars['ICForagerLifespan'] <= 16):
vp_pars['ICForagerLifespan'] = np.random.normal(
loc=float(parameters['ICForagerLifespan_mean']),
scale=float(parameters['ICForagerLifespan_sd']))
vp = VarroaPop(parameters=vp_pars,
weather_file=weather_path,
vrp_file=VRP_FILE,
verbose=False,
unique=True,
keep_files=save,
logs=logs)
vp.run_model()
rep_responses[:, :,
rep] = filter_rep_responses(vp.get_output(),
dates_str=all_dates)
mean_cols = [var[0] + "_mean" for var in RESPONSE_VARS]
sd_cols = [var[0] + "_sd" for var in RESPONSE_VARS]
trt_responses_mean = pd.DataFrame(np.mean(rep_responses, axis=2),
columns=mean_cols)
trt_responses_sd = pd.DataFrame(
np.std(rep_responses, axis=2, ddof=1),
columns=sd_cols) # Note: uses sample sd, not population sd
trt_responses = pd.concat([trt_responses_mean, trt_responses_sd],
axis=1)
#print("Treatment {} responses: {}".format(trt,trt_responses))
start_row = len(DATES) * index
end_row = start_row + len(DATES)
#all_responses.loc[start_row:end_row,:] = trt_responses
all_responses = all_responses.append(trt_responses, ignore_index=True)
#Generate labels for rows and columns
rows = ['_'.join(x) for x in product(TREATMENTS, all_dates)]
response_cols = [x[0] for x in RESPONSE_VARS]
cols = ['_'.join([x, y]) for y in ['mean', 'sd'] for x in response_cols]
all_responses['Index'] = pd.Series(rows) # Add our row labels
all_responses.set_index("Index",
inplace=True) # Set row labels to be the index
#print('Final result: {}'.format(all_responses))
#filtered_resp = all_responses.loc[:,'Adults_mean'] # Keep only the mean number of adults
n_dates = len(all_dates)
return_dfs = {}
for response in RESPONSE_VARS:
filtered_resp = all_responses.loc[:, response[0] + "_mean"]
wide = np.empty([6, n_dates]) # 6 treatments, n_dates days,
for i in range(6):
wide[i, :] = filtered_resp.iloc[i * n_dates:(i + 1) * n_dates]
wide_df = pd.DataFrame(wide,
index=["0", "10", "20", "40", "80", "160"],
columns=all_dates)
return_dfs[response[0]] = wide_df
# add all individuals sum
all_ind_mean = all_responses.loc[:,
all_responses.columns.str.
contains('_mean')].sum(axis=1)
wide = np.empty([6, n_dates]) # 6 treatments, n_dates days,
for i in range(6):
wide[i, :] = all_ind_mean.iloc[i * n_dates:(i + 1) * n_dates]
wide_df = pd.DataFrame(wide,
index=["0", "10", "20", "40", "80", "160"],
columns=all_dates)
return_dfs["All"] = wide_df
return return_dfs
def simulate_all_dates(pars, save = False, logs = False):
"""
Simulate data from the colony study using a set of VarroaPop parameters
This version returns the population at every singel date - can be used to
create posterior predictive plots
:param pars: Dictionary of parameters to vary.
ICQueenStrength_mean
ICQueenStrength_sd
ICForagerLifespan_mean
ICForagerLifespan_sd
AIAdultLD50
AIAdultSlope
AILarvaLD50
AILarvaSlope
:return a dict of arrays of varroapop summary stats(site x day) for adult/pupae/larvae/eggs/all)
"""
start = datetime.datetime.strptime(START_DATES[1], "%m/%d/%Y") #pick earliest start date
end = datetime.datetime.strptime(END_DATE, "%m/%d/%Y")
all_dates = [(start + datetime.timedelta(days=x)).strftime("%m/%d/%Y") for x in range(0, (end - start).days)]
parameters = pars.copy() #copy our inputs so that we don't ever modify them (pyabc needs these)
static_pars = {'SimEnd': END_DATE, 'IPollenTrips': 8, 'INectarTrips': 17, 'RQEnableReQueen': 'false'}
for name, value in parameters.items():
if not name.endswith(('_mean','_sd')):
static_pars[name] = value
static_pars['NecPolFileEnable'] = 'true'
weather_path = os.path.join(DATA_DIR,'external/weather/weather_2015','18815_grid_39.875_lat.wea')
all_responses = pd.DataFrame()
for index, site in enumerate(SITES):
exposure_filename = 'neonic_profile_' + site + '.csv'
exposure_path = os.path.join(DATA_DIR,"processed/neonic_profiles/", exposure_filename)#os.path.abspath(os.path.join('data', exposure_filename))
site_pars = generate_start(static_pars.copy(), site, INITIAL_DF)
site_pars['NecPolFileName'] = exposure_path
site_pars['ICQueenStrength'] = 0
site_pars['ICForagerLifespan'] = 0
while not (1 <=site_pars['ICQueenStrength'] <= 5):
site_pars['ICQueenStrength'] = np.random.normal(loc = float(parameters['ICQueenStrength_mean']), scale = float(parameters['ICQueenStrength_sd']))
while not (4 <= site_pars['ICForagerLifespan'] <= 16):
site_pars['ICForagerLifespan'] = np.random.normal(loc = float(parameters['ICForagerLifespan_mean']), scale = float(parameters['ICForagerLifespan_sd']))
vp = VarroaPop(parameters = site_pars, weather_file = weather_path, vrp_file = VRP_FILE,
verbose=False, unique=True, keep_files=save, logs=logs)
vp.run_model()
site_response = filter_site_responses(vp.get_output(), dates_str= all_dates)
all_responses = all_responses.append(site_response, ignore_index=True)
#Generate labels for rows and columns
rows = ['_'.join(x) for x in product(SITES, all_dates)]
all_responses['Index'] = pd.Series(rows) #Add our row labels
all_responses.set_index("Index", inplace=True) #Set row labels to be the index
n_dates = len(all_dates)
return_dfs = {}
for response in RESPONSE_VARS:
filtered_resp = all_responses.loc[:, response[0]]
wide = np.empty([10, n_dates]) # 10 sites, n_dates days,
for i in range(10):
wide[i, :] = filtered_resp.iloc[i * n_dates:(i + 1) * n_dates]
wide_df = pd.DataFrame(wide, index=SITES,
columns=all_dates)
return_dfs[response[0]] = wide_df
# add all individuals sum
all_ind_mean = all_responses.sum(axis=1)
wide = np.empty([10, n_dates]) # 10 sites, n_dates days,
for i in range(10):
wide[i, :] = all_ind_mean.iloc[i * n_dates:(i + 1) * n_dates]
wide_df = pd.DataFrame(wide, index=SITES,
columns=all_dates)
return_dfs["All"] = wide_df
return return_dfs