10
$\begingroup$

What areas of statistics have been substantially revolutionised in the last 50 years? For example, about 40 years ago, Akaike with colleagues revolutionised the area of statistical model discrimination. About 10 years ago, Hyndman with colleagues revolutionised the area of exponential smoothing. About XX years ago, ...

How do I possibly continue the list, with years and names please? By statistics I mean its all four types from Bartholomew's 1995 presidential address, Chambers's greater and lesser statistics together, as featuring in Hand's recent presidential address on 'Modern statistics' and so on - anything professionally relevant.

$\endgroup$

closed as not constructive by robin girard, user28, user88 Aug 19 '10 at 15:10

As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extended discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance. If this question can be reworded to fit the rules in the help center, please edit the question.

  • $\begingroup$ The only way you can keep this question open is by making it community wiki so please tick the case. $\endgroup$ – robin girard Aug 19 '10 at 7:03
  • 2
    $\begingroup$ However, I have the feeling that this is subjective, argumentative and will require extended discussion please read stats.stackexchange.com/faq I vote to close but encourage you to ask a more specific question (since the idea of the question is good but way too wide). $\endgroup$ – robin girard Aug 19 '10 at 7:05
  • $\begingroup$ one of the extended discussion that could start: are you sure that Prof Rob Hyndman was a reasearcher when parzen and Rozenblatt proposed exponential smoothing :) ? $\endgroup$ – robin girard Aug 19 '10 at 7:08
  • 1
    $\begingroup$ I think with the availability of more powerful computers, different kinds of methods suddenly become practical and important (would one use e.g. boosted decision trees without fast computers ?) $\endgroup$ – Andre Holzner Aug 19 '10 at 7:32
  • 1
    $\begingroup$ Answering the question is not really a clear indication that you would vote to close. People saw my comment, they saw your answer, ... 10 very fast heterogeneous answers in less than an hour ! looks like a chat room ;) $\endgroup$ – robin girard Aug 19 '10 at 13:25

11 Answers 11

15
$\begingroup$

Efron's work on the Bootstrap comes to mind.

$\endgroup$
  • $\begingroup$ like so ;-) $\endgroup$ – user88 Aug 19 '10 at 10:44
15
$\begingroup$

The application of Bayesian statistics with Monte Carlo methods.

$\endgroup$
11
$\begingroup$

Ensemble methods like boosting, bagging, ... etc are another potential candidate.

$\endgroup$
9
$\begingroup$

In 1960 most people doing statistics were calculating with a four-function manual calculator or a slide rule or by hand; mainframe computers were just beginning to run some programs in Algol and Fortran; graphical output devices were rare and crude. Because of these limitations, Bayesian analysis was considered formidably difficult due to the calculations required. Databases were managed on punch cards and computer tape drives limited to a few megabytes. Statistical education focused initially on learning formulas for t-testing and ANOVA. Statistical practice usually did not go beyond such routine hypothesis testing (although some brilliant minds had just begun to exploit computers for deeper analysis, as exemplified by Mosteller & Wallace's book on the Federalist papers, for instance).

I recounted this well-known history as a reminder that all of statistics has undergone a revolution due to the rise and spread of computing power during this last half century, a revolution that has made possible almost every other innovation in statistics during that time (with the notable exception of Tukey's pencil-and-paper EDA methods, as Thylacoleo has already observed).

$\endgroup$
7
$\begingroup$

John Tukey's truly strange idea: exploratory data analysis. http://en.wikipedia.org/wiki/Exploratory_data_analysis

$\endgroup$
6
$\begingroup$

Generalized linear models due to the recently deceased John Nelder and Robert Wedderburn.

$\endgroup$
4
$\begingroup$
$\endgroup$
  • $\begingroup$ What about SAS and spss? $\endgroup$ – Shane Aug 19 '10 at 11:29
  • $\begingroup$ And don't forget Stata. $\endgroup$ – Thylacoleo Aug 19 '10 at 12:38
  • $\begingroup$ Probably deserving of its own question: Which statistical package has made the most revolutionary contribution to the science and practice of data analysis and statistics? $\endgroup$ – Jeromy Anglim Aug 19 '10 at 12:44
  • $\begingroup$ That would be overly argumentative. I think having an answer here that acknowledges all statistical software is on point. $\endgroup$ – Shane Aug 19 '10 at 13:06
  • $\begingroup$ @Shane. Fair enough. I used to use SPSS. Now I use R. R revolutionised the way that I think about and conduct data analysis. It made data analysis fun. I can't speak too much about Stata and SAS, so I'll leave it for others to justify why they might be revolutionary. $\endgroup$ – Jeromy Anglim Aug 19 '10 at 13:32
4
$\begingroup$

There was a great discussion on metaoptimize called "Most Influential Ideas 1995 - 2005" Which holds a great collection of ideas.

The one I mentioned there, and will repeat here, is the "revolution" in the concept of multiple comparisons, specifically the shift from using FWE to FDR methods, for testing very many hypothesis (like in micro array or fMRI and so on)

Here is one of the first articles that introduced this notion to the scientific community: Benjamini, Yoav; Hochberg, Yosef (1995). "Controlling the false discovery rate: a practical and powerful approach to multiple testing". Journal of the Royal Statistical Society

$\endgroup$
  • $\begingroup$ what is FWE and FDR? I suppose FWE to be family wise error, but the other?? $\endgroup$ – Henrik Aug 24 '10 at 11:56
  • $\begingroup$ Well, this thread is subjective, so who knows... Now seriously - FDR stands for false discovery rate (wikipedia it) $\endgroup$ – Tal Galili Aug 24 '10 at 13:45
4
$\begingroup$

Cox proportional hazards survival analysis: http://en.wikipedia.org/wiki/Cox_proportional_hazards_model

$\endgroup$
4
$\begingroup$

The Box-Jenkins approach to time-series modelling: ARIMA models etc.

http://en.wikipedia.org/wiki/Box-Jenkins

$\endgroup$
3
$\begingroup$

The creation of this site ;-)

$\endgroup$
  • $\begingroup$ is turing the site into a discussion forum a revolution ;) ? $\endgroup$ – robin girard Aug 19 '10 at 11:41

Not the answer you're looking for? Browse other questions tagged or ask your own question.