← All research spotlights

Research spotlight · Unit 5

Xiao-Li Meng

Harvard University

Research field
Statistical inference, data quality, missing data, Markov chain Monte Carlo

Xiao-Li Meng is the Whipple V. N. Jones Professor of Statistics at Harvard University and the founding editor of the Harvard Data Science Review. In a 2018 paper he proved that the error of a sample mean from any sample, random or not, is the product of three factors: a data quality measure (the correlation between being recorded and the value recorded), a data quantity measure that depends on the fraction of the population sampled, and the standard deviation of the variable. For self-selected data the first factor does not shrink as the population grows, so the tiny correlation of about 0.005 that he estimated from a 2016 election survey would make a self-selected sample of 2.3 million responses, 1 per cent of all eligible voters, worth no more than a random sample of about 400 people. The unit presents this result as the modern form of the Literary Digest lesson: more data of the same biased kind do not help.

In the masterclass

Section 2, after the Literary Digest and the four kinds of bias, presents his decomposition of the error of a sample mean and the big data paradox it implies.

The research is presented in Unit 5: Sampling and the Central Limit Theorem, in a research spotlight box that explains the result at the level of the unit and links it to what you have just learned.

The paper

Meng, X.-L. (2018). Statistical paradises and paradoxes in big data (I): Law of large populations, big data paradox, and the 2016 US presidential election. The Annals of Applied Statistics, 12(2), 685‑726. https://doi.org/10.1214/18-AOAS1161SF

The scientists named here are not contributors to the Statistics Masterclass, are not affiliated with it, and have not endorsed it. The program presents and explains their published research, with references to the original papers.