meaning of cmp with one equation

[ale dono]

Hi, I am new to the forum. I am using the cmp command to work with random effect model.
However, I am not sure I have got the meaning of it when using only one equation.

For example:

xi: cmp (var1 = var2 var3 var4 i.year || id : ) , ind (1) nolr difficult vce(cluster id)

with 0<var1<1 uncensored
what is it exactly doing?
ind(1) and ind(8) yield the exact same results, so I don't understand what is it doing?

A panel model with time fixed effect and random id effect? A Tobit? or OLS?

Thank you,
Best wishes,
Ale

[David Roodman]

As stated in the documentation, ind(1) means linear, uncensored model. So this command line specifies linear, uncensored model with a random effect at the id level.
As noted in the documentation, a value of 8 in ind() is deprecated; it means that the dependent variable is truncated--except that this command line doesn't include a trunc() option, there is no truncation, and you get the same answer.

The help file contains many examples, including one-equation models with random effects. I suggest you try to understand those.

[ale dono]

thank you for the quick reply, i appreciated. Since I read that "ind(1) = equation is "continuous" for this observation, i.e., has the OLS likelihood or is an uncensored observation in a tobit equation", I am not sure what estimator the routine is using in that specific case

[David Roodman]

It is assuming that for all observations, the difference between the dependent variable its linearly predicted value is normally distributed.

This is also the assumption for uncensored observations in a tobit model.

Given such models, the estimator is Maximum Likelihood.

[ale dono]

one more question: before starting the maximum likelihood iteration, which model is it estimating and showing? I refer to the first table that comes up below the 'for exact fits of equation alone, run cmp separetely on each.' Thank you.

[David Roodman]

In this case, it is fitting ordinary OLS, with no random effect.