3

Today, I read a post Computing and Sustainability: What Can Be Done?. And I found that the author of this post can easily find statistical problems in other fields, such as computer science. Since as a stats graduate, it is very important to recognize statistical questions involved in other fields, I am really curious how to figure out the statistical questions effectively. What are the characteristics for any problems to be recognized as statistical?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Honglang Wang
  • 915
  • 3
  • 9
  • 16
  • 4
    ----- Data ----- –  Jul 02 '12 at 15:39
  • If you got the sequence data of one sample, what kind of statistical problems you could raise? – Honglang Wang Jul 02 '12 at 16:08
  • @Honglang Pretty much any *question* you can ask of data, you can legitimately ask of one value--the real issue is what *answers* can you come up with :-). To see that this is not a trivial point, please read the example of a [one-value confidence interval](http://stats.stackexchange.com/a/1836/). – whuber Jul 02 '12 at 17:00
  • @Momo +1 for a good quip, but aren't you implying there's no difference between statistics, machine learning, or data mining? Do you really mean that? If so, how would you respond to Beth Chance's survey of [definitions of statistical thinking](http://www.amstat.org/publications/jse/v10n3/chance.html) in the JSE (see section 2)? Honglang: a common answer to your question (which Chance quotes) is that statistics is "what a statistician does," so one might say--without providing much illumination!--that a statistical problem is one a statistician is (or ought to be) working on :-). – whuber Jul 02 '12 at 17:04
  • 1
    @whuber Indeed I think that the distinction between ML, DM and Stats to be different fields rather than sides of the same dice is mostly superficial and an unfortunate result of historic processes. But then the fields grow together more and more anyway (hasn't this been discussed elsewhere here?) However, your definition would make the fields being the same anyway but also showering ;-) Although, now that I think of it some mathematical statistics might be done data-free? – Momo Jul 02 '12 at 20:09
  • 1
    @Momo If what you mean by a "field" is its subject matter, then there's good reason to agree that statistics, machine learning, and data mining may be growing together. But by a field, or intellectual discipline, many people mean additional things including the kinds of problems that are deemed important, ways of thinking about them, methods that are developed to attack them, and even the social communities that grow around those fields. In these senses stats, ML, and DM clearly differ. For instance, DM usually does not focus on creating probability models of variation, but stats usually does. – whuber Jul 02 '12 at 20:18
  • 1
    To me data is not enough to make a problem statistical. Suppose the data fit a sine wave perfectly. That may be interesting mathematically but it has no relevance in statistics. What makes a problem statistical is random variability. You may be looking to discover a functional relationship between variables but the relationship is not immediately apparent becaues of the random noise. Through smooth, filtering or averaging statistics allows you to see the signal through the noise. Problems with noisy data are abundant in all fields and that is why statistics is so useful. – Michael R. Chernick Jul 02 '12 at 22:41
  • Could anyone say something together with the post I linked? How could the author of the post be able to say that the computer science problem is a statistical problem? What's the rule behind it to define a statistical problem? – Honglang Wang Jul 03 '12 at 16:06
  • @HonglangWang I think one of the key lines is "In particular, there is a need to develop better statistical models for understanding uncertainty". Modelling uncertainty is an important research line in Probability and Statistics. If you are interested on this, you should take a look at [Spiegelhalter's website](http://www.statslab.cam.ac.uk/Dept/People/Spiegelhalter/davids.html). –  Jul 04 '12 at 08:30
  • The initial link is now broken. Is this the post? -- https://simplystatistics.org/2012/07/02/computing-and-sustainability-what-can-be-done/ – rolando2 Sep 15 '18 at 19:13

0 Answers0