Announcement

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

  • crossed-classified multilevel model syntax

    Dear Statalist,

    I was hoping someone could confirm whether I using a correct syntax for my model.

    My dataset consists of observations that are, on the one hand, nested within CEOs, who are nested within firms, which are nested within industries, whereas on the other, they are also nested within years. This sounds to me like a crossed-classified multilevel model. However, reading the manuals left me uncertain about my syntax, since the examples mostly show how to model data with cross-classification of two levels.

    My syntax looks like:

    Code:
    mixed y x || _all:R.year || industry: || firm:  || ceo:
    I use Stata 14.2.

    Thank you for helping me out.

    Best,
    Miha

  • #2
    I believe your syntax is correct for the model you describe.

    That said, do you really want to use random intercepts for year? I don't know what your data look like, but since you have variables called industry, firm, and ceo, I'm guessing this is finance or economic data. So I'm going to speculate that you don't have very many years represented. If you have only a small number of years, then you will not be adequately sampling the space of random year-effects to get a useful estimate of variance at that level. It might make more sense to just include i.year as a fixed effect in the model.

    Comment


    • #3
      Thanks Clyde!

      My sample spans 18 years and I am doing an age-period-cohort study on CEOs. So I want to compare how coefficients and standard errors of regressors behave when I model period effects in different ways. Indeed, as you suggested I also include years as fixed effects in of the models.

      Comment


      • #4
        I feel like I'm missing something here. Wouldn't each firm have only a single CEO? If so, you don't really have CEOs nested within firms. CEO and firm would be the same level.

        Comment


        • #5
          The dataset was designed to have at least two CEOs per firm.

          Comment

          Working...
          X