st: Cross-classified 3-level model

["Fernando Fernandes de Lima" <fernandopcf@yahoo.com.br>]

Dear Statalists,

I have to model a database, for which my independent variable is the amount
of time spent to make some product. I have different kinds of products which
can be made by different employees. 

Any employee, theoretically, can make any product, although it is more
common to exist some specialization. These employees are grouped in
different geographic areas.

So I have:
Level 1 - product (variable time)
Level 2 - Employee and Kind of product (which perpasses all groups - any
group or employee can make any kind of product).
Level 3 - Group of employees.

In stata I used the following command:

xtmixed time || _all:R.kind || employee: || group:, mle var

Do you think I am correct to use this? If so, when I get the results, I am
interpreting them as:

var(R.kind) - variance due the kind of product
group: var(cons) - variance due to the group of employees
employee: var(cons) - variance due to the employee itself
var (residual) - variance due to the interaction between employee and kind
of product.

Another possibility was to use

xtmixed time || group: || employee:, mle var

In this case I have just:

group: var(cons) - variance due to the group of employees
employee: var(cons) - variance due to the employee itself
var (residual) - variance due to kind of product, since it perpasses all
levels.

Which analysis is correct?

I really appreciate any help,

Fernando Fernandes

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/