Comments on: Data science, big data and statistics
- 161 Downloads
We are delighted that Galeano and Peña give an insightful discussion on the recent change and development of statistical methods driven by big data. Here, we will introduce a different type of example, showing briefly how to combine traditional statistical techniques with ‘machine learning’ models. (Indeed, the latter could be named as ‘modern’ statistical techniques.) This links to the methods discussed in Sections 2.1, 2.2, 2.4, 2.5 and 2.6 in their paper.
The aim of the example is to model the recovery level of stroke rehabilitation patients remotely using accelerometer smartwatch movement data (see, e.g. Noorkõiv et al. 2014). Accelerometer sensors are ideal for this application, because they are cheap and unobtrusive and allow for data collection in naturalistic conditions wherever the user is situated. Since the data are collected over a longer period of time compared to laboratory-based assessments, the data are less likely to exhibit a high degree of observation bias. In...