RESEARCH METHOD COHEN ok

248 INTERNET-BASED RESEARCH AND COMPUTER USAGE
the world may be too complicated for numbers!).
Tymms indicates the limitations of existing school
effectiveness research that is based on linear
premises, however sophisticated. Instead, pouring
cold water on much present school effectiveness
research, he argues:
simulation models would suggest that even if it were
possible to arrange for exactly the same classes to
have exactly the same teacher for two years in the
same classroom living through the same two years
that the outcomes would not be the same.
(Tymms 1996: 132–3)
For him, it is little surprise that school effectiveness
research has failed to account effectively
for variance between schools, because such research
is based on the wrong principles. Rather, he
argues, such variance is the natural outcome of
the interplay of key – common – variables.
In the third example, Gilbert and Troitzsch
(2005: 117–23) report a study of post-war gender
desegregation in German high schools and high
school teachers. The model, using the MIMOSE
program, used 4,500 teachers in 150 schools
of three types, and shows the closeness of the
computer model to the real-life situation observed:
avalidationofthesimulation.Themoduleuses
only three assumptions: first, all the teachers
who leave their jobs are replaced with an equal
probability/opportunity to be chosen, by men and
women (p. 117); second, men remain in their jobs
for twice as long as women; third, new women
take up posts in a individual school with a
given probability which varies according to the
proportion of its women teachers.
This chapter will not discuss the stages of
developing computer simulations (e.g. identifying
the question, defining the target for modelling,
conducting initial observations to establish
the parameters and key features, establishing
the assumptions underpinning the simulation,
verification of the implementation of the
simulation, validation of the simulation (its
correspondence to the real-world situation that
it is modelling), and sensitivity analysis of the
simulation’s responsiveness to initial conditions
and changes to parameters (Gilbert and Troitzsch
2005: 18–19). Nor will it discuss the different
kinds of simulations (e.g. system dynamics,
microsimulation, queuing models, multilevel
models, cellular automata, multi-agent models,
learning models). We refer readers to Gilbert and
Troitzsch (2005) for fuller analyses of computer
simulation and their different types.
Advantages and disadvantages of
computer simulations
Bailey (1994: 322–4) suggests that simulations
have advantages such as:
economy: they are cheaper to run than the
real-life situation
visibility: they can make a phenomenon more
accessible and clear to the researcher
control: the researcher has more control over
the simulation than in the real-life situation
safety: researchers can work on situations that
may be too dangerous, sensitive, ethically
questionable or difficult in real life natural
situations.
Computer simulations are powerful in that,
as well as enabling researchers to predict the
future (e.g. in economic forecasting), simulations
also enable them to understand and explore a
phenomenon. Simulations can act as a substitute
for human expertise, sometimes enabling nonexperts
to conduct research that, prior to
the advent of computers, would have been
the exclusive preserve of experts: Gilbert and
Troitzsch (2005: 5) cite the example of geologists,
chemists and doctors. Gilbert and Troitzsch also
suggest that computer simulations are useful for
training purposes (e.g. pilots) and, indeed, for
entertainment. However, Gilbert and Troitzsch
(2005: 5) underline the prime importance of
computer simulations as being discovery and
formalization of theory (i.e. clarity, coherence,
operationalization, inclusion of elements, and
completeness of a theory).