Three thoughts this time, for our first edition of Thoughts of the Week.
An estimate that is slightly biased but robust, model-independent, easy to compute, and easy to interpret, is better than one that is a non-biased, difficult to compute, mysterious, or not robust. That's one of the differences between data science and statistics.
Learning how to code, especially SQL, should be the last step in becoming a data scientist, because it creates highly rigid thinking, while data science is about ideas, problem solving, vision, initiative and flexibility (the opposite of coding).
Data science is not built on stats taught in traditional stats programs. It is based on modern stats that are not found in stats textbooks, though quite a bit can be found in modern machine learning books.
Source for this sarcastic picture: timoelioot.com