+ Reply to Thread
Results 1 to 4 of 4

Thread: Bidirectional relationship with longitudinal data: what's the best model?

  1. #1

    Bidirectional relationship with longitudinal data: what's the best model?




    N=1000+. I have about 3-4 repetitive measures of a few numerical variables and categorical variables.

    How do I test the hypothesis that x (numerical) increases with y (numerical) and z incidence (categorical) over time, in the same time that both y and z are also affected by the increase in x? In summary, X is fueled by y and z, but both of them are boosted by increase in X over time.

    I cannot say that y and z are independent, right? Do I need to perform 3 different models? 1) x as dependent, 2) y as dependent, 3) z as dependent?

    What would be the best model to test this kind of hypothesis?

    Thanks!

  2. #2
    Points: 3,006, Level: 33
    Level completed: 71%, Points required for next Level: 44

    Posts
    177
    Thanks
    1
    Thanked 29 Times in 29 Posts

    Re: Bidirectional relationship with longitudinal data: what's the best model?

    you want to do an autoregressive model (time series) for a system of equations.

    such that Y = intercept + X + Z + error
    and X = intercept + Y + Z + error

    correct? That's a little bit of a tougher model to work with since its non-recursive.

    You might get more mileage out of something like =
    Y = intercept + X + Z + error
    X = intercept + Z + Error

  3. The Following User Says Thank You to the42up For This Useful Post:

    neuromniscience (03-23-2016)

  4. #3
    Points: 25, Level: 1
    Level completed: 49%, Points required for next Level: 25

    Posts
    3
    Thanks
    3
    Thanked 0 Times in 0 Posts

    Re: Bidirectional relationship with longitudinal data: what's the best model?

    Quote Originally Posted by the42up View Post
    you want to do an autoregressive model (time series) for a system of equations.

    such that Y = intercept + X + Z + error
    and X = intercept + Y + Z + error

    correct? That's a little bit of a tougher model to work with since its non-recursive.

    You might get more mileage out of something like =
    Y = intercept + X + Z + error
    X = intercept + Z + Error
    Thank you for your input. I'll read more about it. I did not think about a forecast model. Can I do a forecast model using repetitive measures for each subject?

  5. #4
    Points: 3,006, Level: 33
    Level completed: 71%, Points required for next Level: 44

    Posts
    177
    Thanks
    1
    Thanked 29 Times in 29 Posts

    Re: Bidirectional relationship with longitudinal data: what's the best model?


    Quote Originally Posted by neuromniscience View Post
    Thank you for your input. I'll read more about it. I did not think about a forecast model. Can I do a forecast model using repetitive measures for each subject?
    yes, there are a number of different approaches depending on what you are asking.

    If you want to know if the system of equations remains the same over time, you test for a stability coefficient.

    if you want to treat the time as a "nuisance" you treat it as psuedo replication in something like a mixed model

    if you want to look at the change over time, you can use a latent growth curve model.

  6. The Following User Says Thank You to the42up For This Useful Post:

    neuromniscience (03-24-2016)

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats