re: st: loneway and intraclass correlation

[David Airey <david.airey@vanderbilt.edu>]

> In Martin Bland's book on medical statistics, he says:
>
> Sometimes we have pairs of observations where there is no obvious  
> choice between X and Y. The choice of X and Y is arbitrary.
>
> He then goes on to discuss intraclass correlation.
>
> I have a set of data which meets this criteria and if I run  
> 'loneway x y' and 'loneway y x', I get 2-different ICC values,  
> which I can expect.
>
> However...what I don't know is what to do with the different  
> values. Is one better than the other? How do I select which ICC  
> value is actually the one I should be using?
>
> Thanks,
> Joe

I don't agree with the premise. I think the intraclass correlation is  
a measure of variation within compared to between classes. One of the  
variables should be continuous (Y) and the other a class or  
categorical variable (X) to use loneway (or xtmixed). In your  
situation, are both your variables continuous measures? To me, that  
would make sense and if you switch either variable as dependent  
variable, you still would get the same correlation, if you used  
Stata's "correlate" command.
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/