Maths EE ideas library

Maths Extended Essay ideas: statistics

Estimation, testing and the mathematics behind statistical methods. The strongest statistics EEs study a method itself rather than run a test on some data.

Statistics ideas (10)

Research questions to start from

One starting question for each idea. Change the case, the comparison or the limit until it is yours.

Why divide by n − 1? Accessible
Why is dividing by n − 1 rather than n needed for an unbiased estimate of variance, and when does the biased estimator actually perform better? A short proof followed by a genuine trade-off (bias against mean squared error) to evaluate.
Where does the least-squares line come from? Accessible
How can the least-squares regression line be derived by calculus and by geometry, and how sensitive is it to outliers compared with alternatives? A method studied from the inside, with an evaluation of its weakness.
How quickly does the central limit theorem kick in? Solid
How large must a sample be for the mean of skewed data to be approximately normal, and how does skewness affect the answer? Turns 'n ≥ 30' into a question you can investigate and argue about.
When combining data reverses the result Accessible
Under what conditions can a trend in every group reverse when the groups are combined, and how can this be predicted algebraically? An algebraic characterisation of a paradox, with real examples you analyse.
Counting animals you cannot see Solid
How accurate is the Lincoln–Petersen estimate of a population size, and how does the Chapman correction reduce its bias? Estimation theory with a real use, plus a simulation you can run at home with beans.
Estimating the largest serial number Solid
What is the best way to estimate the total number N from a sample of serial numbers, and how do the estimators compare? Competing estimators with exact properties you can derive and compare.
Where does the chi-squared test come from? Ambitious
Why does the chi-squared statistic follow approximately a chi-squared distribution, and how small can expected frequencies be before the test fails? A study of a test's foundations and limits rather than a use of it.
Spearman vs Pearson Accessible
When do Spearman's rank and Pearson's correlation coefficients disagree, and which better detects a monotonic but non-linear relationship? Two methods compared with a proof of the Spearman formula and designed test cases.
Can resampling replace formulas? Ambitious
How well do bootstrap confidence intervals for a mean match the formula-based intervals, and when does the bootstrap fail? A modern method examined critically, with clear measures of success.
The mathematics of rating players Solid
How does the Elo rating system turn results into ratings, and how well do its predicted win probabilities match outcomes? A model you can derive, analyse and test on results you collect.

Frequently asked questions

Is statistics a good area for a maths EE?

It can be, if you narrow it to one question you can answer with mathematics you understand. Ask a question about the method (when it works, how it fails, how two methods compare) rather than about a dataset. That keeps the mathematics at the centre.

Can I use one of these research questions as it is?

Use it as a starting point. Change the case, the comparison or the limit until the question is your own, and agree it with your supervisor.

Other areas: Number theory · Algebra · Analysis & calculus · Geometry · Probability · Modelling · Graph theory · Codes & algorithms · Optimisation & games · History of maths