COVID-19 Records

Chat message

Slack / Private Message Drop, p.1037 [SLACK_001251] · slack_pm:msg:10843

Page text: p.1037 · original PDF

Date
2021-12-03 02:30
Type
chat message · slack
recipient
Kristian G. Andersen, Robert F. Garry, Edward C. Holmes
speaker
Andrew Rambaut

Recipients on this medium are inferred from channel membership, not per-message addressing.

Just putting this here first - I am not particularly happy with the some of analysis in the market origins paper. As Ihave said before, a continuous 2D KDE makes no sense for the within market map as it is not a continuous space. Iam also not that comfortable with the use of KDEs in Wuhan. A KDE is an estimate of a continuous probability distribution from which the samples are drawn. The contours are then areas that contain a particular proportion ofthe probability. Saying that the market lies within a particular contour isn't really saying what we are implying it issaying. I think we need to have a discussion about this with Marc, Philippe and Moritz to nail the spatial stats. Orsimply go the route of more of a descriptive narrative - show there are spatial correlations but not try and get technical.

In context

A single line often inverts meaning once you see what it answers, so neighbouring messages are always shown.

  1. 2021-12-03 02:30 Andrew Rambaut
    Just putting this here first - I am not particularly happy with the some of analysis in the market origins paper. As Ihave said before, a continuous 2D KDE makes no sense for the within market map as it is not a continuous space. Iam also not that comfortable with the use of KDEs in Wuhan. A KDE is an estimate of a continuous probability distribution from which the samples are drawn. The contours are then areas that contain a particular proportion ofthe probability. Saying that the market lies within a particular contour isn't really saying what we are implying it issaying. I think we need to have a discussion about this with Marc, Philippe and Moritz to nail the spatial stats. Orsimply go the route of more of a descriptive narrative - show there are spatial correlations but not try and get technical.

Full conversation →