2 or 3 level multilevel model? Quick query

#1
Hi there,

The title should read 3 or 4 level multilevel model, my apologies.

I am estimating a multilevel model (in Stata).
My goal is to examine the effects of the explanatory variables on the country level.
My data is clustered within age cohorts and then countries with years clustered within that. However, my data is measured at the individual level as is common practice. I cannot accurately estimate the effects at the individual level because several explanatory variables are missing at the individual level throughout many years.

My questions are:

1. Since I am not estimating the effects at the individual level as I am missing some explanatory variables, I will not include any individual levels explanatory variables, only country explanatory variables. Is this correct?

2. As the observations are measured at the individual level, do I need to include the country level random effects as well as the cohort random effects and year random effects? i.e. should I have a 3 or 4 level multilevel model?
I thought 4 but if it is 4, (cohort >> nation >> year >> individual)* then my regression results would be explaining the effects at the individual level. Then again, how would it work with a 3 level model (cohort >> nation >> year)** if the observations are measured at the individual level?

*Ex. Stata code 4-level: Meologit y x1 x2 x3 x4 || _all: R.cohort || _all: R.nation || _all: R.year
**Ex. Stata code 3-level: Meologit y x1 x2 x3 x4 || _all: R.cohort || _all: R.year
(x1...x4 explanatory variables measured at the country level; y1 dependent binary variable measured at the individual level)

Any advice would be greatly appreciated.

Many thanks,
 
Last edited: