Advertisement
If you have a new account but are having problems posting or verifying your account, please email us on hello@boards.ie for help. Thanks :)
Hello all! Please ensure that you are posting a new thread or question in the appropriate forum. The Feedback forum is overwhelmed with questions that are having to be moved elsewhere. If you need help to verify your account contact hello@boards.ie
Hi there,
There is an issue with role permissions that is being worked on at the moment.
If you are having trouble with access or permissions on regional forums please post here to get access: https://www.boards.ie/discussion/2058365403/you-do-not-have-permission-for-that#latest

Multinomial Logits

  • 22-03-2010 5:45pm
    #1
    Daithio2
    Closed Accounts Posts: 37 ✭


    Anybody have experience of these? I am trying to model transition rates between 11 different possible response variable states. It is a necessary condition that the individual is already in one of these states to begin with, and the most likely outcome is that the individual won't in fact change state.

    Would it be OK to run one logit, with dummies for the 11 different current states, with 11 potential responses, one of which being that the individual doesn't change? This doesn't feel right to me for some reason.

    Or should I have two regressions. The first being a binary regression with two outcomes, change and don't change, and then have a second regression for all those nominated to change to determine where they end up. Would I have to run 11 different regressions depending on where the individual is coming from?

    Apologies if this isn't very clear, I'm not sure I really understand the issues myself!

    Thanks in advance


Comments

  • #2
    Time Magazine
    Registered Users, Registered Users 2 Posts: 8,452 ✭✭✭Time Magazine


    The dummies thing isn't very good. The sign and coefficient on the x's may well differ as a person changes from 1 to 3 or 2 to 3. Using dummies will, I think, just get the average effect.

    I think your "change or don't change" regression in addition to the eleven logits sounds like a good idea, though I don't know an awful lot about MNLs.


Advertisement