@(@\newcommand{\B}[1]{ {\bf #1} }
\newcommand{\R}[1]{ {\rm #1} }
\newcommand{\W}[1]{ \; #1 \; }@)@This is dismod_at-20221105 documentation: Here is a link to its
current documentation
.
Purpose
Usually the prior for the rate is smooth.
This requires lots of data, at a fine age spacing,
to resolve a jump in a rate at an unknown age.
If the age at which a jump occurs is known, it is possible to resolve
the jump with much less data by specifying a prior that has this knowledge.
Parameters
The following values are used to simulate the data and define the priors
and can be changed:
iota_near_zero
This is the true value of
iota
up to and including age 20.
Note that it is close to zero, but not equal to zero, so that we can
use the rate case
iota_pos_rho_zero
.
iota_after_20
This is the true value of
iota
for ages greater than 20.
iota_eta
Offset in log transformation used for values of eta.
age_table
The age_table
does not need to be monotone increasing.
For this example, it is the same as the table of ages at which
iota
is modeled .
You can changed the order of age_table
and it will not affect the results.
time_table
The time_table
does not need to be monotone increasing.
You can changed the order of time_table
and it will not affect the results.
Model Variables
This example's variables are all
parent rates
for iota
.
The value of
iota
is modeled at each age in the age_table.
The prior for the value of
iota
up to age 20 is a constant equal to
iota_near_zero.
After age 20 it is uniform with lower limit iota_near_zero,
upper limit 1 and mean iota_after_20 / 4.0
(The mean is only used for the initial value and scaling the optimization.)
The prior for the forward age differences in
iota
at age 20
is uniform, and above age 20 it is a Log-Gaussian with mean 0 and
standard deviation 0.1 (about 10 percent coefficient of variation).
Truth
For this example the rate
iota
is constant
with value iota_near_zero for ages less than or equal 20,
and iota_after_20 for ages greater than 20.
Simulated Data
For this example, the simulated data is all
Sincidence
; i.e.
direct measurements of the value of
iota
.
There is no noise simulated with the data; i.e., it is equal to the
'true' value of
iota
.
On the other hand, its is modeled as if there is a 10% coefficient
of variation in the data; i.e., as if there were measurement noise with
standard deviation equal to 10% of the measurement value.
There is a measured value for each age in the age_table
that is greater than 20.