On 1 Jun 2001, Francis Dermot Sweeney wrote:
> Say I have three set of data x,y,z.
I'll assume that you mean "I have three variables, x, y, z."
(One cannot in general perform a regression analysis unless the variables
of interest are in the same data set, not in three separate data sets.)
> x and z consist of data over the time interval (0,T), but for y, there
> is "missing" data, and we only have data on y on (S,T) with 0<S<T.
> I want to perform a least square regression of z onto x and y.
I note with mild interest that you seem not to propose to use time as a
predictor.
> How do I account for the missing data of y?
Depends. If you have theory that specifies how y should behave between
over the time interval (0,S), given its behavior over (S,T) (and/or the
behavior of z and z in that interval), use it to estimate the missing
data. (But be aware that your number of degrees of freedom for error
will be spuriously large, as reported by the regression routine when you
regress z on x and y. You will probably have to make some adjustments
by hand (including the value of the F-test, which depends on the error
mean square, which in turn depends on the number of d.f. for error) to
the results reported by the program.)
Of course, the utility of your results will then depend on how closely
that theory approximates reality.
If you have no reasonable basis for guessing values of y in the interval
(0,S), you cannot "account for the missing values of y". You can carry
out the regression on the interval (S,T), of course; but from the fact
of your question I suppose that would be unsatisfactory for some reason.
-- DFB.
------------------------------------------------------------------------
Donald F. Burrill [EMAIL PROTECTED]
348 Hyde Hall, Plymouth State College, [EMAIL PROTECTED]
MSC #29, Plymouth, NH 03264 603-535-2597
184 Nashua Road, Bedford, NH 03110 603-471-7128
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
