Intra-class correlations (ICC) of a variable

[Viacheslav Iurkov]

Dear all,

I would like to ask whether Stata allows to obtain intraclass correlations (ICC1 and ICC2) for a certain variable. If yes, what is the syntax for doing this?

In one of the papers (relevant to my research), the authors use the ICC analysis to check whether the variable should enter the multilevel model as a random intercept or a random slope. The dependent variable is Return on Assets (ROA), which is observed over years (level 1) over business units (level 2) of the corporations (level 3). If both ICC1 and ICC2 are high, then the authors introduce the variable as stable (random slope) and if ICC1 is low then the variable is introduced as transient (random intercept).

The procedure from the paper is depicted in the attachments.

I would be very grateful for your help.

Attachments:

Model specification

[Phil Bromiley]

If you had bothered to look in the subject index that comes with Stata's documentation, you would have found icc as a procedure.

[Viacheslav Iurkov]

Thank you for the reply. I have bothered. ICC1 is the one-way random-effects model (also can be obtained by using -loneway- command) and ICC2 refers to the two-way random-effects model. But it does not come clear from the instructions how what I ask can be done with the multilevel panel data structure (like the one from the example). Perhaps I should have been more precise when asking the questions.