Notable Advances in Statistics: 2000-2009
During this decade, Functional Data Analysis, a nonparametric statistical technique for which each sample element is considered to be a function, became an established field and textbooks began to appear. LASSO (least absolute shrinkage and selection operator) was popularized in 1996 and the more general Least Angle Regression (LAR) algorithm was proposed in 2004.3
Although random (decision) forests for improved classification and regression were introduced in 1990s, the methodology was developed extensively in the 2000’s. Genetics data analyses, such as microarray analysis and tests for the proportion of multiple alleles, became hot areas for applied statistical research.
In 2002, it was announced that the amount of information stored digitally surpassed non-digital storage. Significance magazine was launched in 2004. John Ioannidis pointed out that, in medical research, most published findings were false.6
It was discovered that the lack of reproducibility of published findings was a wider problem than just biomedicine. Reproducibility became a major scientific and statistical concern. It became clear that science needed to make fundamental changes to its culture. Scientists needed to improve their strategy for scientific discovery and their system for publishing scientific papers. The process for funding science and scientists need to to be reformed. The historical reliance on statistical significance at the 0.05 level received intense criticism.
The American Statistical Association authorized 4 new sections, the Nonparametric Statistics Section (NPAR), the Statistics in Defense and National Security Section (SDNS), the Statistical Learning and Data Science Section (previously called SLDM, the Statistical Learning and Data Mining Section), and the Statistical Programmers and Analysts Section (SSPA).
Last revised: 2021-04-19