3 thoughts on “Infovis vs. statistical graphics: My talk tomorrow (Tues) 1pm at Columbia

  1. Enjoyed your talk today. I sort of thought that the historical comparison charts for baby name endings could have used some kind of normalization (it doesn't change the result, but it seems like it would eliminate any potential distortion).

  2. From your concluding slide

    Data displays as alternatives to photos and cartoons
    Versus
    (Properly calibrated) Discovery of the unexpected

    May explain my attempt to get displays (painting inference) as alternatives to mathematical analysis

    http://www.stat.columbia.edu/~cook/movabletype/ar

    and the recurring concerns about whether noticed patterns in “almost individual observation likelihood” plots can be fully calibrated (i.e. repeatedly good at separating apparent from real outliers)

    K?
    p.s. art show was cancelled due to concerns about possibly upsetting folks …

Comments are closed.