def roi_df(args, model_name, roi):
if args.fixed_t:
args.roi = roi # Temporary
csv = Path(args.data_path) / ("covidtimeseries_%s.csv" % args.roi)
csv = csv.resolve()
assert csv.exists(), "No such csv file: %s" % csv
stan_data, t0 = ncs.get_stan_data(csv, args)
global_start = datetime.strptime('01/22/20', '%m/%d/%y')
frame_start = datetime.strptime(t0, '%m/%d/%y')
day_offset = (frame_start - global_start).days
else:
day_offset = 0
model_path = ncs.get_model_path(args.models_path, model_name)
extension = ['csv', 'pkl'][args.fit_format]
rois = ncs.list_rois(args.fits_path, model_name, extension)
if args.rois:
rois = list(set(rois).intersection(args.rois))
fit_path = ncs.get_fit_path(args.fits_path, model_name, roi)
if args.fit_format == 1:
fit = ncs.load_fit(fit_path, model_path)
stats = ncs.get_waic_and_loo(fit)
samples = fit.to_dataframe()
elif args.fit_format == 0:
samples = ncs.extract_samples(args.fits_path, args.models_path,
model_name, roi, args.fit_format)
stats = ncs.get_waic(samples)
df = ncs.make_table(roi,
samples,
args.params,
stats,
quantiles=args.quantiles,
day_offset=day_offset)
return model_name, roi, df
def roi_df(args, model_name, roi):
model_path = ncs.get_model_path(args.models_path, model_name)
extension = ['csv', 'pkl'][args.fit_format]
rois = ncs.list_rois(args.fits_path, model_name, extension)
if args.rois:
rois = list(set(rois).intersection(args.rois))
fit_path = ncs.get_fit_path(args.fits_path, model_name, roi)
if args.fit_format == 1:
fit = ncs.load_fit(fit_path, model_path)
stats = ncs.get_waic_and_loo(fit)
samples = fit.to_dataframe()
elif args.fit_format == 0:
samples = ncs.extract_samples(args.fits_path, args.models_path,
model_name, roi, args.fit_format)
stats = ncs.get_waic(samples)
df = ncs.make_table(roi,
samples,
args.params,
stats,
quantiles=args.quantiles)
return model_name, roi, df
fit_path = ncs.get_fit_path(args.fits_path, model_name, roi)
df_roi = df_weights.loc[roi]
model_name_weight = model_name + '_weight'
weight = df_roi[model_name_weight]
if args.fit_format == 1:
fit = ncs.load_fit(fit_path, model_path)
# stats = ncs.get_waic_and_loo(fit)
samples = fit.to_dataframe()
samples_weighted_df = samples.sample(frac=weight, replace=True)
dfs.append(samples_weighted_df)
elif args.fit_format == 0:
samples = ncs.extract_samples(args.fits_path, args.models_path,
model_name, roi, args.fit_format)
stats = ncs.get_waic(samples)
samples = fit.to_dataframe()
samples_weighted_df = samples.sample(frac=weight, replace=True)
dfs.append(samples_weighted_df)
df_model_averaged = pd.concat(dfs)
df_model_averaged.reset_index(inplace=True, drop=True)
df_model_averaged.to_csv(Path(args.fits_path) / f'DiscreteAverage_{roi}.csv')
# now that we have model averaged fits, create tables
# mimic other tables code and merge reweighted table with model averaged table
# replace applicable regions with model averaged results
# Get all model_names, roi combinations
combos = []
extension = ['csv', 'pkl'][0] # use 'csv'
rois = ncs.list_rois(Path(args.fits_path), 'DiscreteAverage', extension)
combos += [('Discrete1', roi) for roi in rois]