Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • How to normalise dissimilarity and exposure indices (for longitudinal data)

    Hi everyone!
    I'm currently working with longitudinal data and I wanted to compare school and residential segregation (natives vs minorities) over time and across specific territorial sub-units (NUTS 3 level).
    The idea is to compute dissimilarity and exposure indices in each territorial unit using schools as obs. But I've noticed that the number of schools per territorial units are not fixed over time. It might have happened that some schools have been closed or merged with others.
    Since the number of schools are not fixed, this might bias my results since "there is less space to move". Therefore, I wanted to ask you if there are methods to compute the D and E indices in a "normalised" way so to take into consideration the non-fixed n of schools over time.

  • #2
    I think this is the wrong way round. In this situation it does seem possible that your indices vary with number of schools as well as with other predictors and so checking that is part of the needed analysis.

    Any kind of pre-analysis adjustment raises more problems than it solves, including

    how to do it -- how to work out a procedure and assess it analytically or through simulations

    a side issue becoming a bigger deal than it deserves

    losing comparability with other studies.


    Comment


    • #3
      Cross-posted at https://stats.stackexchange.com/ques...gitudinal-data

      Please note our policy on cross-posting, which is that you are asked to tell us about it. https://www.statalist.org/forums/help#crossposting

      Comment

      Working...
      X