Syntax of cross-classified multilevel model

[Eunhee Kim]

Dear Statalist,
I need to analyze data with complex structures.

My data set consists of individuals who are fully nested within teams, which are nested "crossly" within functions and locations, which are fully nested within firms.
(individuals) ---> (teams) ---> (functions) and (locations) ---> (firms)

I guess I need to do four-level cross-classified multi-level analysis and I have the following hypothetical syntax:

Code:
xtmixed y x || firm: || _all:R.function: || _all:R.location: || team:

I hope someone can confirm whether it is right, or correct it if it is wrong. Thank you for helping me out.

Best,
Eunhee

[Joseph Coveney]

What's "nested 'crossly' within functions and locations" mean?

Unless there's some reason to consider function and location as random, I would probably treat them as fixed and go for an easier-to-fit model.

mixed y c.x i.function##i.location || firm: || team:

Depending upon the research question and the nature of the phenomenon, I'm not even sure that team would be there, either.

[Eunhee Kim]

Thanks Joseph,
The "nested crossly" means that teams are nested within both functions and locations, but functions and locations are not nested within each other.
The hypothetical data structure looks like this:

ind	team	function	location	org	other predictors....
1	1	1	1	1	...
2	1	1	2	1	....
3	1	2	2	1	...
4	1	2	3	1	...
5	1	3	3	1	...
6	2	1	2	1	...
7	2	1	2	1	...
8	2	1	3	1	...
9	2	2	3	1	...
10	3	4	4	2	...
11	3	4	5	2	...
...	...	...	...	...	...

And I need to consider function and location as random and there are multiple predictors in each level.
For example, the dependent variable is individual performance (individual level), and predictors are
: individual personalities (individual level)
: team characteristics such as team diversity, team leader personalities, and team structures (team level)
: functional characteristics such as the importance of each function, number of required tasks that teams of each function needs to perform (function level)
: location characteristics such as whether it is a big city or not, the customer population of each location (location level)
: organizational characteristics such as firm performance, firm culture and climates

Is there any way to analyze this kind of data?
Thank you for your help!

[Joseph Coveney]

I'm not sure why you feel the need to consider function and location as random. In my experience, fitting models with cross-classified random effects by iterative maximum likelihood methods works best when the data are much richer than what your table implies for your case. So, I still recommend handling those two predictors as fixed. (I'm not sure that your dataset can even support a function × location interaction term in the fixed-effects equation, let alone support them as crossed in the random-effects side.)

.mixed y c.x i.function i.location i.personality i.diversity . . . || firm: || team: