[Replying to OP and to list. -- DFB.]
Short answers embedded below.
On Thu, 25 Mar 2004, Anupama Ramachandran wrote:
> I have a utility model in which time is treated as a dummy variable
> with values 0 and 1.
Until corrected, I will assume that what you actually mean is "a
dichotomous variable which is coded {0,1} for convenience". As Paul
Swank pointed out, "indicator variable" is sometimes used for such a
variable. But OK so far...
> In the utility model, time will interact with at least two other
> terms.
Sounds perfectly reasonable.
> If I specify time to be a dummy variable are interactions allowed
> with dummy variables?
Yes. Equivalent to any two-level factor in ANOVA contexts, or any
dichotomous variable in MR contexts.
> What are the specifics I should follow while doing this?
Too general a question to answer. More context (and in general more
information) would help. What, for instance, do you have in mind by
"while doing this"? What kind of formal model are you trying to fit?
("Utility model" does not convey much to me, although it probably does
to an econometrician.)
> Are there any papers which deal with the treatment of time as a dummy
> variable?
Scads. In education or psychology, one often has pretest/posttest data.
The pre/post paradigm may not be readily applicable in your domain of
discourse -- as Paul pointed out, it would be important to know whether
your observational units are crossed with "time" or nested within it,
and for most pre/post studies they're crossed.
(Whether a dichotomous variable is coded {0,1} or {-1,+1} or {1,2} --
or any other two values, and I have sometimes found {6,13} useful -- the
coding has no effect on how the variable operates in the model one is
using (or building), except in such minor details as the actual value of
(say) a regression coefficient (but it will not affect any statistical
test associated with that coefficient). Or in generating apparent but
spurious multicollinearity among the predictors and thir interactions --
see my White Paper "Modelling and interpreting interactions in multple
regression" on the Minitab web site www.minitab.com, for an example with
one continuous and three dichotomous predictors.
> I'd really appreciate any help or suggestions in the same.
------------------------------------------------------------
Donald F. Burrill [EMAIL PROTECTED]
56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
