Simulating Probabilities |
||

Main Concepts | Demonstration | Activity | Teaching Tips | Data Collection & Analysis | Practice Questions | Milestone | Fathom Tutorial | ||

Teaching Tips • Beware: don't lose the point of the simulation. It's about estimating a probability, not performing an intricate series of taps on a calculator. • It is tempting to use simulations -- particularly computer
based simulations -- to help students "discover" a concept, or to help
them understand a concept. This is trickier than it seems. If you are
not careful, students will see the simulation as a blackbox. • It is best to do hands-on simulations first, before computer simulations, so that students can "get their hands dirty" for example using a die or a deck of cards to represent a random experiment. • Students will not learn by running other people's simulation programs. They must do their own simulations. • It's very important for students to understand exactly what is happening at each step of a simulation exercise. • Students must do many different simulations before the idea "sinks" in. • Don't confuse lower case n -- the number of objects in a
sample in a single trial -- with N, the number of times you repeat the
trial in your simulation. For example, a trial might consist of
selecting 100 Californians at random and counting the number who voted
for Arnold Schwarzenegger, under the assumption that 60% of the
population supported him. To simulate these, we have a bag of six red
chips and four black chips. We draw 100 chips with replacement and
count the number of reds. We repeat this 5000 times. In this example, n
= 100 and N = 5000. (No one said simulations were easy!) • There are three types of probabilities mentioned that
students sometimes confuse: theoretical probabilities (made using
mathematical arguments), experimental probabilities (based on data),
and simulated probabilities that come from simulation exercises and are
used to approximate theoretical probabilities. • There is no rule for deciding when your approximation of the theoretical probability is "good enough." "Good enough" depends on the context and the accuracy required of the context. • Some students may wonder what it means for trials to be independent. In the context of a simulation, independent trials means the outcome of any one trial does not affect our assessment of the probability distribution for the outcomes of any other trials. • Galton built a famous mechanical simulation: the quincunx.
This simulation illustrates the Central Limit Theorem for the binomial
distribution and could also be used to calculate approximate
probabilities involving adding a series of yes/no random outcomes. You
can view a computer version of it here: RAND
Central Limit Theorem in Action. There are many other quincux
displays, but this is the best. • A classic probability problem that can be fun to simulate with your students is the birthday problem (assuming you have over 30 students in your class.) One possible moral of this problem is that coincidences are more likely than one might think. Which leads us to think that sometimes what we perceive as meaningful patterns are actually simply due to chance. |