On Tue, 04 Feb 2003 17:13:10 +0100, Mark <[EMAIL PROTECTED]> wrote: > Hi all, > > Doing tests on a production function, I apply a trend correction first > and then calculate the logarithmic data. Would it be wise to do this in > the reverse order, calculating the log data first and making the log > series stationary? > > I am using a standard Cobb-Douglas production function.
I can't say that I know what you are talking about, but (a) Does it make any difference? - what fractions of change are involved? Is (max/min) on the order of 1.1 -- in which case, it hardly matters -- or 10? (b) My usual assumption is that measurement properties belong to the raw measure, so the transformation comes early. That also tends to simplify descriptions, and keep other options open.... -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
