SS-18 ICBM
Oscillator
Would it be accurate to say that statistics as a distinct branch of mathematics for analyzing characteristics of data only achieved this form with the invention of computers?
History questions not worth their own thread V
- Thread starter Plotinus
- Start date
- Status
No? People did statistics (e.g. error bars on measurements) before WWII, I suppose. And automated statistics programs are pretty recent, I suppose.
Probably did the same thing as astronomers: Have calculators, namely women doing the calculations for you.
Probably did the same thing as astronomers: Have calculators, namely women doing the calculations for you.
SS-18 ICBM
Oscillator
There were computers before the integrated circuit.
Yes, and there were statisticians before even these. A cursory glance at Wikipedia's timeline, and my own experience from being taught statistics as a physicist would lead me to suggest that statistics as a branch of mathematics fledged in the 18th century.
Key figures like Gauss, Bayes, Bernoulli and Laplace spring to mind.
SS-18 ICBM
Oscillator
What kinds of data sets and samples were they analyzing then?
madviking
north american scum
I remember reading in 'The Ascent of Money' by Niall Ferguson that the first statisticians, as evidenced in the Wiki article that you linked, were insurance guys.
Niall Ferguson is wont to refer to things by weird labels. For example, he also refers to Christianity as a "killer app". Him referring to these people as "statisticians" does not mean that they actually were.
madviking
north american scum
My words, actually, not his. I just remember vaguely reading about it. I should actually go and find that passage in the book but it's ever so slightly out of reach from my chair.
Well, their experiments?
I would guess that the scale and complexity have increased drmatically. Because one computer can do far more in a given time frame than a hundred people. And, given it is properly programmed, there is a significantly reduced risk of error in the calculations.
TheLastOne36
Deity
- Joined
- Jan 17, 2007
- Messages
- 14,045
subscription post. 
It still all depends on data collected. Comparable ata from before the "Modern" era grows progressively more scarce the further one moves back. Furthermore, data needs to be comparable, i.e. being recorded according to the same procedure. Given the trouble to compare modern crime statistics, such universality of procedure is even today severely lacking.
SamSniped
DJ Goodboye
Random question, but was St. Patrick an Italian by heritage? I remember reading somewhere that he was the son of Roman parents, but grew up in Roman-controlled Britain.
He was of Romano-British stock; this probably means his family were Britons who had long adopted a Roman way of life.
Traitorfish
The Tighnahulish Kid
Well, that depends on what part of Briton he was from, and on most of the Western seaboard the emphasis fell on "Briton" rather than "Roman".
That is true; the other way of looking at it is that his family were of Roman descent but had lived in Britain for sufficiently long that they were considered 'provincials': you could probably compare 'Irish-American' here for the variety of meanings.
"Briton" being a meaningful term in the fourth century?
sophie
Break My Heart
Traitorfish
The Tighnahulish Kid
If you'd find yourself capable of expanding on that cryptic comment, I'm sure we'd all be grateful. I make no pretence of being a classicist, I won't be offended.
MCdread
Couldn't she get drowned?
- Joined
- Jun 21, 2001
- Messages
- 5,348
Most of the key and basic concepts of statistics are way older than that. If you root it with probability theory, then it kicked off in the late 17th/early 18th centuries with people like Pascal and others, who were trying to describe mathematically games of cards and dice.
If you want to focus on the "canonical" distributions, then the Gaussian distribution was developed ca. 200 years ago (together with the least squares method; the central limit theorem was developed during the 19th century). I reckon it was discovered either in the context of astronomical observations or it found immediate applicability there (although it soon became obvious it could describe all sorts of phenomena).
The binomial distribution is probably even older. Ditto for the Poisson distribution which was famously found to describe the distribution of accidental deaths by horse kicks in the 19th century Prussian army. Others, like the t-student and the gamma distribiutions are early 20th century discoveries.
In fact, by the 1920s and 1930s, you could say frequentist statistics (the one most people use) was a mature science. The most influential guy was Fisher who was developing work in population genetics (or at least, that's what he became famous for). Most statistical tests, p-values and whatnot date from this period.
Bayesian statistics on the other hand also started in the 18th/19th centuries. It was a bit of a niche thing for many decades, but it is now reemerging triumphant over over frequentist statistics (at least, amongst physicists, computer scientists, etc. It will take a while to spread out to fields that are not so quantitative-minded). I guess what computers provided, in this and other areas, was a fast and powerful means of implementing numerical algorithms for sampling from all kinds of distributions and conducting more sophisticated tests. This allowed statisticians to deal with types and sizes of data (and problems) that would have been impossible to tackle before.
At the same time, it is fair to say that the existence of the computer stimulated other developments that would probably not have happened otherwise. Machine learning would be the obvious example, as the name itself implies. Even though pre-computer methods such as principal component analysis can be considered a form of machine learning, modern machine learning and inference (the kind of stuff that data scientists use every day) would either be unknown or just the obscure pet subject of a handful of isolated academics in the absence of modern computers imo.
Finally, statistical mechanics stemmed out of thermodynamics in the late 19th century, with guys like Boltzmann, Maxwell, Gibbs, etc. Theirs and latter ideas are now used everywhere, from bioinformatics to finance. Computers were obviously great instruments for popularizing said ideas and models. An important link was later established with information theory, which is of huge importance in an inclusive definition of modern statistics (or things that modern statistics uses). Information theory has been developed since the 1940s in the field of telecommunications. Its ideas and applications are independent of whether we need or have access to computers. Computers probably did solidify and transform the field though.
So, all in all, imo the answer to the initial question is both yes and no. Modern statistics would be nowhere near where it is without computers, but the majority of its basic concepts predate computers and were indeed already widely used before that.
- Status
