2016-05-19 Follow-up meeting (2)

[andkov]

Notes during the meeting

Graciela

• (-1) is the first state at the first occasion.
• the only thing that is used is the state at the certain age.

• don't bother adding a missing state for the baseline state

  Iva: We know s died, but don't know the age?

  • you can add the state at the last time they were alive
  • use the last observed state
  • remove observation that does not have the date of

  Ardo on reverse transitions:

  • state 3 = lowest mmse observed so far
  • decision to be made: misclassification model OR re-define state 3
  • misclassification is "smoothing" the data

  Graciela:

  • misclassification seems to be making more sense
  • Marioni used misclassification
  • misclassification seems more realistic
  • I've applied for Whitehall II access.

  Meeting in September:

  • Meeting in Edinburgh instead of Amsterdam

  Resolutions:

  • don't add a missing state for the baseline state; remove the first (missing observation) and move the waves one to the left

  Ardo will simpulate data and devolop the script:

  • fit a 4-state model that allows misclassification for 2, 3