Database Reference
In-Depth Information
some occasions occur later, on others earlier. Also, their severity varies. Can you
explain why this is so?
The results in Figure 2.3 are for five model runs. We have checked the “Compar-
ative” box in the graph pad to enable multiple runs to be displayed in the same page
of the graph pad. Do you expect the model results for a large number of runs—on
average—yield the same pattern as we observed in the previous section?
If you wish to let a parameter vary from run to run along a normal distribution
with known mean and standard deviation, choose the built-in function NORMAL
and specify a mean and standard deviation. Other specifications of random variables
are available as well. Make use of them as you see fit.
Both the RANDOM and NORMAL built-ins allow for the option to specify a
“seed,” which ensures that STELLA uses for subsequent runs the same sequence of
random numbers as it does the first time the model is run with a seed. For example,
if you have two parameters that are specified as random numbers, say A and B, and
you specify
A
=
RANDOM
(
0
,
1
,
1
)
(2.3)
B
=
RANDOM
(
0
,
1
,
2
)
(2.4)
then subsequent runs of the model use the “first” strings of random numbers (all
between zero and one) generated for A and the “second” string of random numbers
(also between zero and one) generated for B. If you do not specify a seed, then all
subsequent runs will differ simply because different random numbers will be cho-
sen. While this may be informative in some cases, as for Figure 2.3, it is sometimes
difficult to see whether changes in a model's results stem the mere fact that the
model uses different random numbers or whether they stem from the fact that the
model itself was changed.
EPIDEMIC WITH RANDOMNESS
CONTAGIOUS(t)
=
CONTAGIOUS(t
−
dt)
+
(GET SICK
−
STAY IN BED) * dt
INIT CONTAGIOUS
=
1
{
Individuals
}
INFLOWS:
GET SICK
=
CONTACT RATE * (CONTAGIOUS
+
SICK) * NON IMMUNE
{
Individuals per Time Period
}
OUTFLOWS:
STAY IN BED
=
CONTAGIOUS
{
Individuals per Time Period
}
IMMUNE(t)
=
IMMUNE(t
−
dt)
+
(RECOVER) * dt
INIT IMMUNE
=
0
{
Individuals
}
INFLOWS:
RECOVER
=
.9*SICK
{
Individuals per Time Period
}
NON IMMUNE(t)
=
NON IMMUNE(t
−
dt)
+
(BIRTHS
−
GET SICK) * dt
INIT NON IMMUNE
=
1000000
{
Individuals
}
INFLOWS:
BIRTHS
=
5000
{
Individuals per Time Period
}
