Dear Kelvyn

It depends on the interest of the Study but I use to present estimates of parameters for element and for cluster of elements too. I think that this is necesary when the cluster is a natural one.

What to report depends somewhat on your research question, but mostly on the conventions of the field. It's unfortunate when convention dictates the answer, though. For instance, what is the average of a male and female? If your subjects are male and female and there is a gender effect, who cares what the average effect is? It wouldn't represent a single person in the study, so why report it?

Dear Kelvyn,

Particularly I prefer to use the cluster when the research involves a set of variables or criteria which act at the same time (multivariate analysis) or present a strong influence among them. Nevertheless, I do not reject the use of a average as an analysis starting point, as well as a good box plot graph to ilustrate data's "behavior". Actually I've been using this combination to highlight some features of the variables in macro and micro scenarios and this approach provides good insights for further discussions. Hope It helps.

Best regards,

Glauco Gomes

Glauco

Thank you - I was asking about a specific technical type of modelling eg longitudinal analysis where the response variable is discrete.

Kelvyn