Announcement

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

  • Dependent Variable with different combinations of items and 2 possible values (0.1)

    Hai all,

    I would like to use a dependent variable that is a combination of health risks happening together. (Smoking , Drinking , Drug , Obesity , Physical Inactivity , Insufficient Nutrient Intake , Sexual Behavior , Insufficient Sleep ). All these variables could have two possible values (0.1) which implies either no smoking or smoking and so on. Thus i want to look at the imoact of some variables on the co-occurence of a particular combination of risk behaviors. How can i implement this.

  • #2
    One practical problem you will run into is that with 9 risky behaviors you have \(2^9=512\) possible pairs. If you also consider groups of three the number becomes even larger. Even if you have enough data to reliably estimate such a model (which I very much doubt) you would end up with way to many parameters to meaningfully interpret and communicate to your audience. So the model needs to be dramatically simplified.

    Possible solutions would be to look at a smaller subset of risk factors, e.g. just look at smoking, drinking, drugs, as they substantively belong together. Alternatively you could look at the number of risk factors, rather than the specific combination.
    ---------------------------------
    Maarten L. Buis
    University of Konstanz
    Department of history and sociology
    box 40
    78457 Konstanz
    Germany
    http://www.maartenbuis.nl
    ---------------------------------

    Comment


    • #3
      Thanks for the remark. The two existing solutions are the ones that is already used in other papers.Total number of risks for each person or look at three factors or do separate regressions for each risk outcome. But i would like to contribute in terms of using more than 3 risk behaviors. The data set that i am using is ALSPAC. Also the idea was to look at the impact of mothers pre-natal lifestyle (High-Medium-Low risk ) , SES and other covariates on multiple risk behaviors in teenagers.

      Specific combination is better in a way that policies could be targeted for that specific risk behaviors. Also is it possible for me to run regressions for (1) Smoking , Drinking , Drug
      (2) Sexual activity , Drinking , Obesity .In a way , i am picking up different combinations on my own which could be supported on the basis of correlations , maybe.

      Thank you

      Comment


      • #4
        I understand why you want to do this, but that does not change the basic fact that there are 512 possible twoway combinations when you have 9 binary variables.

        You might get some leeway with loglinear models (as in models for cross tabulations, not models with log transformed dependent variable). You would have to think long and hard about plausible constraints and on how to present your results.
        ---------------------------------
        Maarten L. Buis
        University of Konstanz
        Department of history and sociology
        box 40
        78457 Konstanz
        Germany
        http://www.maartenbuis.nl
        ---------------------------------

        Comment

        Working...
        X