Gábor’s mantra on research with data

Doing analysis with data

1. Know thy data. Origin stories matter.
2. Garbage data + simple stats = garbage. Garbage data + AI = convincing garbage .
3. Data does not speak. The analyst speaks via data and analysis.

Methods

1. RCT with N=32 « observation study w N== 100k « RCT with 1K.
2. RCT is often not doable for ethical or practical reasosns.
3. Observational analysis is hard but good data and 100+ years of statistics and econometrics can get you close to a causal interpretation

On acquiring knowledge we can act on

1. Our enemies (after sugary drinks) are p-hacking, publication bias, outright cheating driven by researchers and industry interests.
2. Despite best intentions we do not know many things, ok to change view in light if new evidence.