2. What is a model?
A model is a collection of objects that interact to
create a unified whole, such as cell collection
system, a rat or a human.
The type of models that are of interest are
mathematical models that represents the system
of interest.
3. The art of modeling:
Models are built using experimentally derived
data so called data generating process is
dependent on system inputs, system dynamics
and the device used to measure the output from
a system. In addition to this systematic
processes are the sources of error the confound
our measurements.
DATA=SYSTEMATIC COMPONENTS+ERROR
4. output
input System dynamics Exogenous
measurement
system
Measured
data
Process noise
&/or modeling
nosie
Measurment
errors
5. Types of models :
Models can be classified into many different
categories.
Using the nomenclature of Distefalo and Landaw
, pharmacokinetics models can generally can be
broken into two types:
Models of data
Models of system
6. Models of data
Usually refered to as emperical models ,
required few assumptions about the data
generating mechanisms .
These models are useful when little is known
about the underlying physical process from
which the data are generated.
empirical models may be useful at prediction
they should not be extrapolated .
Eg: Allometric scaling
7. Models of systems
Also called as mechanistic models which are
based on physical and physiological principles
and should have as many features of the system
incorporated into the model as the data allow.
These models usually take the form of
differential equations or partial differentianal
equations based on mass – balance ,product –
precursor ,or mass –action principles.
Eg: transport into and out of tissues as a function
of blood flow and permeability.
8. Types of models of systems
Time-variant vs time – in variant
Deterministic vs stochastic
Static vs dynamic
Lamped vs distributed
Linear vs non-linear
Continuous vs discrete
9. Properties of useful model
Ability to characterize the observed data and to
include the most important features of the data.
Make accurate mix and precise prediction .
Increases understanding of the system.
The model is actually used .
The model is completed on time.
Logically consistent , plausible.
Validated by empirical obeservations.
10. Cont…
Robust to small changes in the data.
Appropriate level of precision and detail.
As simple as possible.
Judged on what it is intended to do .
Has flexibility .
Is effective as a communication tool.
Severs many different purposes.
may allow for extrapolation out side the data range.
11. The model development process:
Analyse the
problem
Perform
experiment
Collect and
Clean the
data
Formulate
model
Fit the
model
Check
model
Interpret &
Communicate
results
Validate
model
12. Model validation
It is a contentious area. Modelers in the field
cannot even agree to call it validation.
One group uses the term model qualification
instead of model validation, the idea being that a
model can never truly be validated so using the
term validation would be technically incorrect.
A fundamental principle of modeling is that a
model can never be proven, only disproved.
Model validation attempts to disprove a model by
applying a series of validity tests to a model and
its predictions.
13. Cont…
The more validity tests a model passes, the greater
credibility the model will have.
If a model fails a test then the model is disproven
and often times it is back to the drawing board.
The degree of model validation will ultimately
depend on the model objectives.
If the objective is too simply characterize a set of
data, model validation is usually minimal, but if the
objective is prediction,then validation is usually
more comprehensive.
14. Cont..
Definition of model validation refers to the
assessment and extent to which a model is founded
and fulfills the purpose for which it was developed.
A model may also be valid for one set of data, but
not for another.
A model may be valid for one set of assumptions
but not valid for another set.
15. Cont..
Mc Leod reports that a valid model is based on the
following properties:
Degree to which model duplicates past system
behaviour using historic data as an input.
Degree to which model behaviour conforms to
exsiting theory.
Degree to which model accuratly forecasts the
future.
Degree to which the model is found accaptable to
other modelers.
16. Cont…
Degree to which the model is found to be
acceptable to those will use it .
Degree to which the model yields opposite
results when opposite values are assigned to the
values and opposite relationships are postulated
and opposite assumptions are made
17. Guidelines presented by McLeod
Face validated : refers to model being
appropriate on face value.
Credibility : model in which there is a faith or a
belief in the validity predictions or out comes.
Internal validity: refers to a model it self being
judged without regared to external criteria.
External validity: refers to accepets related to out
side model.
pragmatic validity: refers to the model being
able to meet its stated objective in a timely
manner.