xtmixed, multilevel data with ordinal and nominal variables

[peter]

Hi all

My data shows quite complex characteristics.
a) the data is on two levels, decisions nested in firms.
b) my dependent variable is an ordinal variable, theoretically ranging
from 1 .. 14
c) I have many ordinal and nominal independent variables, both on the
level of decisions and on the level of firms.
d) the design is unbalanced, for some firms, I have 1 decision, for
some firms I have 5 decisions.
e) sample is small (110 decision, 40 firms)

I'm quite at ease with normal regression and SPSS, but only recently
started with Stata and am quite a novice with hierarchical data.

I learnt that xtmixed would be an appropriate procedure, since I was
told to make things easier, I shall treat my dependent variable as
interval. Thus, I could use a syntax along the lines of xtmixed x dec1
dec2 dec3 firm4 firm5 firm6, || firm_id:

I have the following questions to the group:
1. Is it a big error if I use my ordinal data as interval? (I know,
gllamm could handle this, but the syntax is too complex for me at the
moment.

2. In SPSS, I can tell the program, which of the independent variables
are factors (i.e. ordinal variables) and what are covariates
(interval). I cannot do this in Stata, does STATA automatically know
what are ordinal and what are interval variables? Apparently, the
results differ very much, in STATA more predictors are significant.

3. In SPSS, I can use the EM-Means function. So if for instance,
Firm_Industry is a
nominal variable, I can investigate, which of the industries are
different from each others. Can I do this in Stata as well?

4. On what basis do I need to decide, whether I should on top of fixed
effects, also include some variables as random? Do I need any random
effects at all with my structure?

5. What is important when interpreting the output (apart from the
significant effects, and the -2ll?)

6. Would I better use xtreg/xtgee, and use Firm_ID as a panel variable?

I know, many questions, but hope someone can help me. I appreciate any
comment. Many thanks in advance!!