Multi-Scale Ecology

QQ Plot Challenge

Look at a randomly generated QQ plot and decide whether the data are reasonably consistent with a normal distribution.

The game includes normal and non-normal distributions and explains the generating distribution after each answer.

• Play QQ Plot Challenge

---

Scientific Literature Experiment

How much of the published scientific literature is actually true? In this experiment you will create your own scientific literature by testing 200 hypotheses.

You decide how many scientific ideas are actually correct, the statistical power of the experiments, and the α threshold used to decide what gets published. You can then see how many true and false findings make it into the published literature.

The experiment demonstrates an important, but sometimes counter-intuitive, point: even when scientists use a seemingly strict significance threshold, the fraction of published findings that are true also depends strongly on how many of the hypotheses being tested were plausible in the first place.

• Run the Literature Experiment

This experiment is inspired by the framework introduced by John Ioannidis in Why Most Published Research Findings Are False, PLOS Medicine (2005). Read the paper .

Statistical Decisions

A statistical error is not simply a number in a textbook. Someone, or something, may carry the consequences.

Work through a series of real-world decisions. First identify the Type I and Type II errors. Then decide which mistake you consider more serious, and what that means for the balance between α, β and statistical power.

There is a correct statistical classification of the errors. There is not necessarily one correct answer to which error society should be more willing to accept.

• Make the Decisions

More games

Future exercises can cover correlation, distributions, regression, confidence intervals, hypothesis testing and statistical decision making.