--------------------------------------------------
lines 5-121 of file: example/user/plot_data_fit.py
--------------------------------------------------

# {xrst_begin user_plot_data_fit.py}
# {xrst_comment_ch #}
#
# Example Plotting The Data and Its Fit
# #####################################
#
# Nodes
# *****
# There are four nodes in this example.
# The world node has one child, north_america.
# The north_america node has two children, united_states and canada.
# The :ref:`parent_node<option_table@Parent Node>` is canada which
# does not have any children.
#
# Rates
# *****
# There is a parent smoothing the *iota* , *rho*
# and *chi* rates.
# This smoothing has one grid point; i.e., the rates are constant
# in age and time.
# There is no child node smoothing so there are no random effects
# for these rates.
# In addition, there is no parent smoothing for the other rates
# so they are zero.
# The value priors for the rate smoothing is uniform with lower limit 1e-4
# and upper limit 1.0. The mean 0.1, is only used as a starting point
# for the optimization.
# There are no difference priors because the smoothing
# has only one grid point.
#
# Integrands
# **********
# The integrands for this example are
# :ref:`avg_integrand@Integrand, I_i(a,t)@Sincidence` ,
# :ref:`avg_integrand@Integrand, I_i(a,t)@remission` ,
# :ref:`avg_integrand@Integrand, I_i(a,t)@mtexcess` , and
# :ref:`avg_integrand@Integrand, I_i(a,t)@prevalence` .
# The first three integrands are direct measurements of the following rates:
# {xrst_spell_off}
# {xrst_code py}
integrand2rate = {
   'Sincidence':  'iota'   ,
   'remission':   'rho'    ,
   'mtexcess':    'chi'    ,
}
# {xrst_code}
# {xrst_spell_on}
# Prevalence is in the integrand table, but it has no simulated data.
# This shows what happens when there is no data for one of the
# integrands in the :ref:`plot_data_fit@integrand_list` .
#
# Data
# ****
# All of the data corresponds to canada.
#
# n_data
# ======
# There are *n_data* data points for each of the integrands where
# {xrst_spell_off}
# {xrst_code py}
n_data  = 100
# {xrst_code}
# {xrst_spell_on}
#
# max_plot
# ========
# This is the maximum number of data points to plot per integrand.
# The points are chosen randomly, but there order is preserved.
# {xrst_spell_off}
# {xrst_code py}
max_plot = int( n_data / 2 )
# {xrst_code}
# {xrst_spell_on}
#
# Measurement Noise
# =================
# The data is simulated a Gaussian with mean equal to the
# corresponding *rate_true* :
# {xrst_spell_off}
# {xrst_code py}
rate_true = {
   'iota'  : 1e-2 ,
   'rho'   : 1e-3 ,
   'chi'   : 1e-4 ,
}
# {xrst_code}
# {xrst_spell_on}
# and with coefficient of variation *meas_cv* :
# {xrst_spell_off}
# {xrst_code py}
meas_cv = 0.2
# {xrst_code}
# {xrst_spell_on}
#
# hold_out
# ========
# There is also one outlier for each integrand
# (that is not simulated with the Gaussian noise described above).
# The outlier is placed in the middle of the data set for each integrand.
# The :ref:`data_table@hold_out` for the outliers is set to one
# (it is zero for all the other data).
#
# Call to plot_data_fit
# *********************
# {xrst_literal
#     BEGIN call plot_data_fit
#     END call plot_data_fit
# }
#
# Source Code
# ***********
# {xrst_literal
#     BEGIN PYTHON
#     END PYTHON
# }
#
# {xrst_end user_plot_data_fit.py}