وقتی داده های کوچک از پس کلان داده بر می آید When small data beats big data

1. Introduction Big data is justifiably a major focus of research and public interest. Even so, small data is still with us. The same technological and societal forces which have generated big data have also generated a much larger number of small datasets. At first glance, more data would seem to be clearly better than less data. All things being equal, this is true. In practice, obtaining more data will involve additional costs of various kinds and will complicate the analysis. In the real world of fixed budgets, there are trade offs between quality and quantity. Sometimes small data will beat big data and reach the right conclusions faster, more reliably and at lower cost. In this article, we present a variety of situations where small data will be preferable. For related discussion in this same special issue, see Secchi (2018). 2. Wider Meaning of Big and Small data The term “big data” means different things to different people. Statisticians 15 tend to think of “big” in terms of size, either many cases or many variables or both. Yet the term has taken on a wider meaning to the public with “big” also refering to the extent, impact and mindshare of the phenomenom. Statisticians have had to adapt their communication to this wider definition. This is now accepted and understood. It is perhaps less well-known among statisticians that “small data” also has a wider meaning in the business community as a reaction to big data. As with big data, the definition proves elusive but we attempt a contrast. Big data deals with the large, observational and machine analysed. Small data results from the experimental or intentionally collected data of a human scale where the focus is on causation and understanding rather 25 than prediction. See the book, “Small Data”, by Lindstrom (2016). Given the hype surrounding big data in the business world, it is refreshing to see some recognition for the virtues of small data.