Pack the punch
Various posts across multiple blogs/websites described Econometrics as a tool for causality identification and explanatory modelling and Machine Learning has been projected as a gizmo for predictive modelling. Well, that is half-truth. Other than explanatory modelling, Econometrics also have remarkable power for forecasting. Timeseries econometrics, a branch of applied econometrics, has achieved repeated success in generating reliable forecasts.
It seems, long stay of econometrics in academia put multiple intellectual layer which snowballed to a gigantic discipline which cannot easily sneak into the doors of many data science practitioners.
In business, the fundamental purpose of data is forecasting. To do this appropriately Econometrics Learning, a hybrid (Econometrics + Machine Learning) discipline should work best. Econometrics and Machine Learning are different and complementary. They should not drift apart but combined to pack a punch. Combination forecasting techniques found to be capable in overcoming inaccuracy generated by any single forecasting model.
Hal Varian, the Chief Economist at Google Inc. listed the new tricks (basically machine learning techniques) that econometrics should adopt. He presented this at Stanford University. The link to the presentation is given below.
The challenge is to simply Econometrics Learning. Unless simplified, its required acceptance among practitioners and users is a distance thought.
What if it cannot be simplified, Plan B; Econometric Learning has to demonstrate achievement. By doing so, it may motivate corporates to adopt it as a specialized domain more like IT, legal, etc. The specialized domains are highly respected and extensively utilized to add considerable value to the overall business.
It seems, long stay of econometrics in academia put multiple intellectual layer which snowballed to a gigantic discipline which cannot easily sneak into the doors of many data science practitioners.
In business, the fundamental purpose of data is forecasting. To do this appropriately Econometrics Learning, a hybrid (Econometrics + Machine Learning) discipline should work best. Econometrics and Machine Learning are different and complementary. They should not drift apart but combined to pack a punch. Combination forecasting techniques found to be capable in overcoming inaccuracy generated by any single forecasting model.
Hal Varian, the Chief Economist at Google Inc. listed the new tricks (basically machine learning techniques) that econometrics should adopt. He presented this at Stanford University. The link to the presentation is given below.
The challenge is to simply Econometrics Learning. Unless simplified, its required acceptance among practitioners and users is a distance thought.
What if it cannot be simplified, Plan B; Econometric Learning has to demonstrate achievement. By doing so, it may motivate corporates to adopt it as a specialized domain more like IT, legal, etc. The specialized domains are highly respected and extensively utilized to add considerable value to the overall business.
Big Data: New Tricks for Econometrics: Hal Varian
http://web.stanford.edu/class/ee380/Abstracts/140129-slides-Machine-Learning-and-Econometrics.pdf
http://web.stanford.edu/class/ee380/Abstracts/140129-slides-Machine-Learning-and-Econometrics.pdf
Machine learning or econometrics?
https://www.linkedin.com/pulse/machine-learning-econometrics-chris-kuo/