Subscribe to DSC Newsletter

Building a house and building a data-analytic model

If I want to build a house, wouldn't it be wise to learn carpentry? Does the analogy hold for data-analytic multivariate models? Or is it simply enough to let a machine do it, with no knowledge by the machine operator of how to interpret the results from those modeling efforts? Or is it true, as one person has recently asserted, that he could replicate ALL statistical procedures and techniques using MapReduce, without knowing anything about statistics and probability, or the vast collection of discipline-specific applications of statistical science in economics, the social sciences, the physical sciences, (including physics and chemistry itself), business or organizational management, archaeology, anthropology, and other historical sciences (evolutionary biology and genetics), and biostatistics, to name a few? Will machine learning supplant all of these careful developed approaches to problems that are peculiar or particular to a very large array of efforts aimed at scientific advance?  

Views: 899

Tags: (the, MapReduce, academic, algorithmic, analysis, and, approaches, auto-didacticism, coding, community), More…credentials, data, it, learn, learning, machine, method, of, probability, professional, scientific, statistical, the, theory, versus, yourself-ism

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Comment by George Brunner on May 2, 2014 at 1:00pm

The answer is "E" all the above. Statistics is not going away nor should it. But if you can combine Machine Learning/ data mining with Statistics and advance exploratory data visualization you have a toolset that can answer most if not every question imaginable. I have seen situations where visualizing stats as simple as variance and correlation coefficients,  make insights jump off the page. But if I can have my cake and eat it too I will. Why not use all the tools in the tool kit and see if you end up with the same answer. In some circles that's called validation, which is always a really good idea when you don't want to be proven wrong by someone using a different approach or toolset. If the answer is write it should be the same no matter how you arrived at it.

Comment by D. L. von Kleeck on May 2, 2014 at 5:30am

It is my opinion that machine learning will not "supplant all of these careful developed approaches", but machine learning is a very powerful discovery tool that can lead to new hypotheses "that are peculiar or particular to a very large array of efforts aimed at scientific advance".

Videos

  • Add Videos
  • View All

© 2019   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service