Simulation
Abstract
Simulation involves calculating exact data for a chosen model
which can then be displayed as a graph or surface to observe
the behaviour of the model over selected ranges of independent
variables given a set of parameter values.
The model may be described by a simple equation or it may be
defined by a nonlinear equation or differential equation which
can only be solved numerically. Special models such as those
required to describe flow cytometry data usually call for
dedicated techniques.
Details
SIMFIT has a number of programs that can be used to simulate deterministic
models or stochastic processes.
usermod define a set of user-supplied models for plotting,
simulating or fitting as an ASCII text file
makdat simulate data (u = f(x), v = g(x,y) or w = h(x,y,z))
deqsol solve sets of nonlinear differential equations
adderr add random error to simulate experiments
rannum random numbers and walks using chosen distributions
makcsa generate flow cytometry profiles
Program MAKDAT
MAKDAT comes along with a library of models, hierarchies of
models and differential equations. You can choose models and
parameter values required and calculate y = f(x) for a range
X_start =< x =< X_stop, either by fixing X_start, X_stop, or
by choosing Y_start = f(X_start), Y_stop = f(X_stop), and then
allowing MAKDAT to find appropriate values for X_start and
X-stop. You have to provide starting estimates for X_start,
X-stop in order to use the second method, so this means that
you have to understand the mathematics of the model.
Advice
With complicated models or differential equations,
fix X_start and X_stop and observe how the graph of
y = f(x) changes as you change parameters and/or
end points. When you have a good idea where your end
points lie you can then try the option to fix y and
calculate X_start and X_stop. This is needed when
f(x) is not monotonic in the range of interest.
The output file from MAKDAT contains exact data for
input into program ADDERR to add random errors.
Program ADDERR
The output file from MAKDAT contains exact data for y = f(x)
which is useful for generating high resolution graphics and
for data simulation.
You may want to add random errors to exact data, to simulate
experimental error. To do this, the output file then becomes
an input file for ADDERR. After adding this random error the
input file is left unchanged (with the exact data) and a new
output file is produced with the perturbed data.
In this way experimental data can be simulated repeatedly.
Model => MAKDAT => Exact data => ADDERR => Simulated data
There are numerous ways to use ADDERR, including generating
replicates. If in doubt pick 7% constant relative error with
3 replicates as this mimics many situations. Note:- constant
relative errors cannot be used where your exact data set has
elements with y = 0.
Back to Help Menu or End Help