1. Refined the first paper: the key difference between standard multinomial logit (individual characteristics matter), conditional logit (alternatives' attributes matter), and mixed model (both matter). Also in technical terms: CLM is conditional on the sum of positive (or "failure" in medical term) outcome, MNL is unconditioned. CLM lends itself to fixed effects treatment, MNL doesn't. Independence from irrelevant alternatives matter, but there is no way to test it in the context of two options only. Problem in RP dataset: our ASC predicts the outcome perfectly.

2. Got a "public warning" from my little sister in Tokyo. Alright, Sis, it's 2004.

3. Music: Dave Brubeck (p) with Paul Desmond (as), Eugene Wright (b), and Joe Morello (d). Recommendation: "Take Five" (Original. Definitely superior than Al Jarreau's vocal version) and "Blue Rondo" (Turkish tone?)