st: Re: ANOVA with random effect

["Joseph Coveney" <stajc2@gmail.com>]

Ada Keding wrote:

I would like to run an ANOVA with a random operator effect.

In my data I have two treatments (A or B, fixed effect) for which a
continuous outcome is measured. However, outcomes are clustered by a
random effect operator. Each operator has one outcome per treatment
group.

Treatment    Outcome   Operator
(fixed)                               (random)
A                     x.xx             1
B                     x.xx             1
A                     x.xx             2
B                     x.xx             2
...                     ...               ...

I tried something like:

anova outcome treatment /  operator|treatment

However, running this, the Residual is 0. I think this model only
applies if I had multiple observations in each treatment arm for each
operator.

Is anyone able to advise how I could get a fixed treatment estimate
while accounting for random operators in the anova command?

--------------------------------------------------------------------------------

It looks like a conventional balanced Randomized Blocks ANOVA, and so it would
be

    anova outcome treatment operator

or, equivalently, 

    xtreg outcome i.treatment, i(operator) fe
    contrast operator

Joseph Coveney


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/