2.17 Unit 2 summary
2.17.1 You should be able to
Use Monte Carlo simulation to explore patterns in a random process:
- Model: Translate real life phenomena into a model to be used in the simulation process.
- Simulate: Use TinkerPlots™ to generate random outcomes from a model.
- Evaluate: Determine the “typical” result from a Monte Carlo model, and a range of likely results
Conduct a statistical hypothesis test of an observed result, including:
- Write an appropriate null hypothesis that specifies a “no effect” probability model and a source of variation
- Use Monte Carlo simulation in TinkerPlots™ to simulate a study if the null hypothesis were true
- Model: Use a sampler to model the study if the null hypothesis were true
- Simulate: Run the simulation hundreds of times and collect the result of interest.
- Find a range of likely results if the null hypothesis were true
- Determine whether the observed result is compatible with the null hypothesis
- Calculate a p-value
- Formulate a conclusion
2.17.2 You should understand
The logic behind statistical hypothesis testing, including:
- Regularity in randomness
- The role of the null hypothesis as specifying a baseline to compare the observed result to
- Why we use Monte Carlo simulation
- Why we need to run multiple trials and when we have run enough
- What the distribution of results represents
- What we are checking for, in order to determine whether the observed result is compatible with the null model
- What a p-value represents
- The sort of conclusions we can (and can’t) make from a statistical hypothesis test.
2.17.3 TinkerPlots™ skills
- Create a new sampler and use different devices to model a null hypothesis
- Plot values from a table and organize (by separating) the the plotted values.
- Numerically summarize values in a plot (e.g.,
- Automatically collect the results from many trials.
- Use the
Dividertools to count the values in a distribution that are as or more extreme than a given value
- Monte Carlo simulation
- Hypothesis test
- Null hypothesis
- “No effect” probability model
- Statistical significance