In this talk, I begin with stating the most fundamental question in data analysis. To outline its resolution and to illuminate where data analysis can and should go, I look at data analysis from the perspective of “element of information”, then argue the most basic form of “a piece of information” in data. The collective of information pieces is further demonstrated to reveal a characteristic phenomenon of complex system: broken symmetry. Showing such a phenomenon through binary bipartite networks in the process of discovering knowledge in several real examples, I argue for the critical need of involving domain scientists in data analysis. Along each step of this knowledge discovering process, I turn the spotlight on incapability of statistical modeling on data.

 

[Current background]

The successful influences of Statistics in many scientific fields are rather evident. However, such successes would not be enough to sustain or justify Statistics as an independent discipline. Since those data analyses are primarily carried out by domain scientists themselves, not by statisticians. While in the field of Statistics, statisticians’ justifications of their works are not based on real scientific evidence, but on seemingly fancy addresses of “mathematical theorems”, which are hardly realistic in real world. Lacking real-world justifications, Statistics (or data Science) will be severely dented by Machine Learning and AI in the foreseeable future. Can it survive?

 

Meeting number (access code): 2511 761 9832

Meeting password: bfSF3zmAB87