1
$\begingroup$

I am conducting a longitudinal dyadic analysis and I thought it is appropriate to treat time clustered within persons, and persons clustered within couples? In David Kenny's book, he mentioned that time is not nested but crossed with persons. Can anyone help me explain what he actually means by this? "In over-time data from dyads, we have three factors: time, person, and dyad. Researchers often make the mistake of considering these data to be a three-level nested model in which time points are nested within persons and persons are nested within dyads. The problem is that time and person are usually crossed, not nested. That is, for a given dyad, the level of time is the same for the two persons at each time point. If the three-level nested model is mistakenly assumed, then the correlation between the two partner’s intercepts, rcc, is constrained to be positive (because it is estimated as a variance), and the correlation between the two members’ errors at each time, ree, is assumed to be zero."

$\endgroup$
  • $\begingroup$ The text does as excellent job in explaining the issue. Which part do you need help on? $\endgroup$ – Michael M May 2 at 9:04
  • $\begingroup$ Hi Michael, what i dont understand is why the level of time is the same for the two persons at each time point. Is it because we usually measure them at the same time?Also, I dont quite understand why the correlation between 2 members's errors at each time is zero if we use 3 level model? should it be the other way? because the errors would be zero if the time level is the same across 2 members? $\endgroup$ – Statsnerd May 3 at 1:01
  • $\begingroup$ This post addresses your question, not for dyads, but for traditional panel data where time is indexed by year or month or whatever. stats.stackexchange.com/questions/439337/… $\endgroup$ – Erik Ruzek May 3 at 21:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.