def store_results(dm, area, sex, year):
types_to_plot = 'p i r rr'.split()
graphics.plot_convergence_diag(dm.vars)
pl.clf()
for i, t in enumerate(types_to_plot):
pl.subplot(len(types_to_plot), 1, i + 1)
graphics.plot_data_bars(dm.model.get_data(t))
pl.plot(range(101),
dm.emp_priors[t, 'mu'],
linestyle='dashed',
color='grey',
label='Emp. Prior',
linewidth=3)
pl.plot(range(101), dm.true[t], 'b-', label='Truth', linewidth=3)
pl.plot(range(101),
dm.posteriors[t].mean(0),
'r-',
label='Estimate',
linewidth=3)
pl.errorbar(range(101),
dm.posteriors[t].mean(0),
yerr=1.96 * dm.posteriors[t].std(0),
fmt='r-',
linewidth=1,
capsize=0)
pl.ylabel(t)
graphics.expand_axis()
pl.legend(loc=(0., -.95), fancybox=True, shadow=True)
pl.subplots_adjust(hspace=0, left=.1, right=.95, bottom=.2, top=.95)
pl.xlabel('Age (Years)')
pl.show()
model = dm
model.mu = pandas.DataFrame()
for t in types_to_plot:
model.mu = model.mu.append(pandas.DataFrame(
dict(true=dm.true[t],
mu_pred=dm.posteriors[t].mean(0),
sigma_pred=dm.posteriors[t].std(0))),
ignore_index=True)
data_simulation.add_quality_metrics(model.mu)
print '\nparam prediction bias: %.5f, MARE: %.3f, coverage: %.2f' % (
model.mu['abs_err'].mean(),
pl.median(pl.absolute(
model.mu['rel_err'].dropna())), model.mu['covered?'].mean())
print
data_simulation.initialize_results(model)
data_simulation.add_to_results(model, 'mu')
data_simulation.finalize_results(model)
print model.results
return model
def store_results(dm, area, sex, year):
types_to_plot = 'p i r rr'.split()
graphics.plot_convergence_diag(dm.vars)
pl.clf()
for i, t in enumerate(types_to_plot):
pl.subplot(len(types_to_plot), 1, i+1)
graphics.plot_data_bars(dm.model.get_data(t))
pl.plot(range(101), dm.emp_priors[t, 'mu'], linestyle='dashed', color='grey', label='Emp. Prior', linewidth=3)
pl.plot(range(101), dm.true[t], 'b-', label='Truth', linewidth=3)
pl.plot(range(101), dm.posteriors[t].mean(0), 'r-', label='Estimate', linewidth=3)
pl.errorbar(range(101), dm.posteriors[t].mean(0), yerr=1.96*dm.posteriors[t].std(0), fmt='r-', linewidth=1, capsize=0)
pl.ylabel(t)
graphics.expand_axis()
pl.legend(loc=(0.,-.95), fancybox=True, shadow=True)
pl.subplots_adjust(hspace=0, left=.1, right=.95, bottom=.2, top=.95)
pl.xlabel('Age (Years)')
pl.show()
model = dm
model.mu = pandas.DataFrame()
for t in types_to_plot:
model.mu = model.mu.append(pandas.DataFrame(dict(true=dm.true[t],
mu_pred=dm.posteriors[t].mean(0),
sigma_pred=dm.posteriors[t].std(0))),
ignore_index=True)
data_simulation.add_quality_metrics(model.mu)
print '\nparam prediction bias: %.5f, MARE: %.3f, coverage: %.2f' % (model.mu['abs_err'].mean(),
pl.median(pl.absolute(model.mu['rel_err'].dropna())),
model.mu['covered?'].mean())
print
data_simulation.initialize_results(model)
data_simulation.add_to_results(model, 'mu')
data_simulation.finalize_results(model)
print model.results
return model
### @export 'data'
#model = data.load('models/af')
model = data.load('/home/j/Project/dismod/output/dm-39544')
model.input_data['age_end'] += 1 # change year-end to preferred format
### @export 'plot-prevalence-data'
df = model.input_data
df = df[df['data_type'] == 'p'] # select prevalence data
df = df[df['area'] == 'USA']
#df = df[df['year_start'] <= 2000]
pl.figure(**book_graphics.three_quarter_page_params)
pl.subplot(1, 2, 1)
graphics.plot_data_bars(df)
pl.xticks()
pl.yticks()
pl.xlabel('Age (years)')
pl.ylabel('Prevalence (per 1)')
pl.axis([-2, 102, -.01, .22])
pl.subplots_adjust(left=.23 + .14, right=.23 + .87, bottom=.14, top=.87)
pl.savefig('book/graphics/af_ages_intervals.pdf')
print 'The systematic review of the descriptive epidemiology of atrial fibrilation included $%d$ observations of disease prevalence for the United States' % len(
df)
### @export 'save-results'
pl.show()
reload(book_graphics)
# <codecell>
book_graphics.set_font()
### @export 'initialize'
df = pandas.read_csv('/home/j/Project/dismod/gbd/data/ssas_mx.csv', index_col=None)
df['age_end'] += 1
# <codecell>
### @export 'initial-rates'
pl.figure(**book_graphics.quarter_page_params)
graphics.plot_data_bars(df)
pl.semilogy([0], [.1], '-')
pl.ylabel('Rate (Per 1)')#
pl.yticks(fontsize='large')
pl.xlabel('Age (Years)')
pl.xticks()
pl.subplots_adjust(.1, .175, .98, .875, .275)
pl.axis([-5, 105, 2.e-4, .8])
min_mx = df['value'].min()
age_min_mx = df['age_start'][df['value'].argmin()]
max_mx = df['value'].max()
age_max_mx = df['age_start'][df['value'].argmax()]
agm.fit_age_standardizing_model(model)
pl.figure(figsize=(17., 11), dpi=72)
pl.xlabel('Age (Years)', size=32)
pl.ylabel('Rate (Per PY)', size=32)
pl.xticks(size=28)
pl.yticks(size=28)
pl.grid()
pl.plot(model.ages, model.pi_age_true, '-', label='Truth', linewidth=10, color=colors[0], zorder=-2)
pl.axis([-5, 105, 0., 1.])
pl.savefig('/media/windows/t/age_group_truth.png')
graphics.plot_data_bars(model.input_data, 'talk')
pl.axis([-5, 105, 0., 1.])
pl.savefig('/media/windows/t/age_group_data.png')
pl.plot(model.ages, model.vars['mu_age'].stats()['mean'], 'k-', label='Posterior mean', linewidth=10, zorder=10)
pl.axis([-5, 105, 0., 1.])
pl.savefig('/media/windows/t/age_group_standardize_0.png')
pl.plot(model.ages, model.vars['mu_age'].trace().T, '-', color='grey', alpha=.1, linewidth=10, zorder=-1)
pl.plot(model.ages, model.vars['mu_age'].stats()['95% HPD interval'][:,0], 'k-', label='95% HPD interval', linewidth=2, zorder=10)
pl.plot(model.ages, model.vars['mu_age'].stats()['95% HPD interval'][:,1], 'k-', linewidth=2, zorder=10)
pl.legend(fancybox=True, shadow=True, loc='upper right', prop={'size':32})
pl.ylabel('Rate (Per PY)', size=32)
pl.xticks(size=28)
pl.yticks(size=28)
pl.grid()
pl.plot(model.ages,
model.pi_age_true,
'-',
label='Truth',
linewidth=10,
color=colors[0],
zorder=-2)
pl.axis([-5, 105, 0., 1.])
pl.savefig('/media/windows/t/age_group_truth.png')
graphics.plot_data_bars(model.input_data, 'talk')
pl.axis([-5, 105, 0., 1.])
pl.savefig('/media/windows/t/age_group_data.png')
pl.plot(model.ages,
model.vars['mu_age'].stats()['mean'],
'k-',
label='Posterior mean',
linewidth=10,
zorder=10)
pl.axis([-5, 105, 0., 1.])
pl.savefig('/media/windows/t/age_group_standardize_0.png')
pl.plot(model.ages,
model.vars['mu_age'].trace().T,
'-',
