def plot_data_and_prediction(data,
means,
stds,
numos,
ax=None,
bins=None,
n_curves=50,
group='x'):
assert ax is not None
ax.hist(data,
bins=bins,
rwidth=0.5,
facecolor='r',
edgecolor='none',
normed=True)
if bins is not None:
if hasattr(bins, '__len__'):
xmin = bins[0]
xmax = bins[-1]
else:
xmin = np.min(data)
xmax = np.max(data)
n_samps = len(means)
idxs = map(int, np.round(np.random.uniform(size=n_curves) * n_samps))
x = np.linspace(xmin, xmax, 100)
ax.set_xlabel('y')
ax.set_ylabel('p(y)')
for i in idxs:
m = means[i]
s = stds[i]
lam = 1 / s**2
numo = numos[i]
nu = numo + 1
v = np.exp([noncentral_t_like(xi, m, lam, nu) for xi in x])
ax.plot(x, v, color=pretty_blue, zorder=-10)
ax.text(0.8,
0.95,
'$\mathrm{N}_{%s}=%d$' % (
group,
len(data),
),
transform=ax.transAxes,
horizontalalignment='left',
verticalalignment='top')
ax.xaxis.set_major_locator(mticker.MaxNLocator(nbins=4))
ax.yaxis.set_major_locator(mticker.MaxNLocator(nbins=4))
ax.set_title('Data Group %s w. Post. Pred.' % (group, ))
def draw_distribution(self, group, x, i):
"""Draw the ith sample distribution from the model, and compute its values for each element of x.
:param string group: specify group, control or variant
:param numpy.array x: linspace vector, for which to compute probabilities
:param int i: index of the distribution to compute
:return: values of the model for the ith distribution
:rtype: numpy.array
"""
m = self.stochastics[group + '_mean'].trace()[i]
s = self.stochastics[group + '_sigma'].trace()[i]
nu = self.stochastics[group + '_nu_minus_one'].trace()[i] + 1
lam = 1 / s**2
return np.exp([noncentral_t_like(xi, m, lam, nu) for xi in x])
def plot_data_and_prediction( data, means, stds, numos, ax=None, bins=None,
n_curves=50, group='x'):
assert ax is not None
ax.hist( data, bins=bins, rwidth=0.5,
facecolor='r', edgecolor='none', normed=True )
if bins is not None:
if hasattr(bins,'__len__'):
xmin = bins[0]
xmax = bins[-1]
else:
xmin = np.min(data)
xmax = np.max(data)
n_samps = len(means)
idxs = map(int,np.round( np.random.uniform(size=n_curves)*n_samps ))
x = np.linspace(xmin, xmax, 100)
ax.set_xlabel('y')
ax.set_ylabel('p(y)')
for i in idxs:
m = means[i]
s = stds[i]
lam = 1/s**2
numo = numos[i]
nu = numo+1
v = np.exp([noncentral_t_like(xi,m,lam,nu) for xi in x])
ax.plot(x,v, color=pretty_blue, zorder=-10)
ax.text(0.8,0.95,'$\mathrm{N}_{%s}=%d$'%( group, len(data), ),
transform=ax.transAxes,
horizontalalignment='left',
verticalalignment='top'
)
ax.xaxis.set_major_locator( mticker.MaxNLocator(nbins=4) )
ax.yaxis.set_major_locator( mticker.MaxNLocator(nbins=4) )
ax.set_title('Data Group %s w. Post. Pred.'%(group,))
def plot_data_and_prediction( data,
means,
stds,
numos,
ax,
bins = None,
n_curves = 50,
group = 'x',
name = '',
colour = 'red',
plot_y = False
):
if colour == 'red':
# for colour plots with data in red use:
light_blue = '#89d1ea'
red = '#FF0000'
elif colour == 'blue':
# for colour plots with data in blue use:
light_blue = '#FF0000'
red = '#89d1ea'
else:
# for greyscale plots use:
light_blue = '#909090'
red = '#101010'
# plot histogram of data
ax.hist(data, bins = bins, rwidth = 0.7,
facecolor = red, edgecolor = 'none', normed = True)
if bins is not None:
if hasattr(bins,'__len__'):
xmin = bins[0]
xmax = bins[-1]
else:
xmin = np.min(data)
xmax = np.max(data)
n_samps = len(means)
idxs = map(int, np.round(np.random.uniform(size = n_curves)*n_samps))
x = np.linspace(xmin, xmax, 100)
if plot_y:
ax.set_xlabel('y', verticalalignment = 'center')
ax.set_ylabel('p(y)')
for i in idxs:
m = means[i]
s = stds[i]
lam = 1/s**2
numo = numos[i]
nu = numo+1
v = np.exp([noncentral_t_like(xi, m, lam, nu) for xi in x])
ax.plot(x, v, color = light_blue, zorder = -10)
ax.text(0.99,0.95,'$\mathrm{N}_{%s}= %d$' % (group, len(data),),
transform = ax.transAxes,
horizontalalignment = 'right',
verticalalignment = 'top'
)
ax.xaxis.set_major_locator(mticker.MaxNLocator(nbins = 4))
ax.yaxis.set_major_locator(mticker.MaxNLocator(nbins = 4))
ax.set_title(name + ' data \nwith Predicted Posteriors',
size = 12, weight = 'bold')
