On Tue, 06 May 2008 19:41:59 GMT, "reflex" <sdfs at (no spam) sdfsd.com> wrote:
can anyone help me with this question: 'what are the advantages of
multi-variate analysis over bivariate analysis?'
You have to put several variables into an equation
(predicting, or being predicted)
if you want to get out some result is a previously
unspecified *pattern* of contrasts.
Clearly... You can't see patterns if you take variables one
at a time... Either way (say, multiple regression or separate
testing with Bonferroni correction) may have approximately
the same power.
If you know what the pattern is that you will get, or
that you are primarily interested in getting, (most often,
everything "good" going the same direction), then
it is much more powerful to *test* that contrast all
by itself. -- If the test is what you are interested in.
Of course, many tyros forget an important first step
either for testing or for understanding is to temporarily
throw away all but the three or four "best" expected
predictors, and draw your *main* conclusions based on
your main interests. (Use the other variables for
confirmation and for further explication.)
all I can think is that it is able to consider as many potentially
important
variables as possible. are there any others?
And you have to put A, B, and C into the equation if
you want to see the effect of D "while controlling for
A, B, and C."
--
Rich Ulrich
http://www.pitt.edu/~wpilib/index.html
