Date:         Thu, 4 Dec 1997 20:14:12 GMT
Reply-To:     Richard F Ulrich <wpilib+@PITT.EDU>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         Richard F Ulrich <wpilib+@PITT.EDU>
Organization: University of Pittsburgh
Subject:      Re: correlation coefficient

William Dudley (wdudley@SPH.EMORY.EDU) wrote:
: >At 13:59 04/12/97 +0000, Safa Gurcan wrote:
: >>Hi,
: >>Hi,
: >>I have 30 chicken, each chicken has repeated measures(not equal) and
: >>2 variables (live weight and egg weight). In general total n size is
: >>392. I want to calculate to correlation coefficient between 2
: >>variables. If i calculate the arithmetic mean of 2 variables for each
: >>chicken, n size will be reduced to 30. I think that if i do this
: >>process a lot of information will be lost.
 
 << snip, details; recommendations >>
 
: I wonder what other readers would do with these data?
 
 
If there is just one live-weight per chicken, then there is only
the possibility of a "between-chicken" sort of correlation or
regression.  For any statistic between-chicken, the n=30; it is not
"reduced" to 30, because there were never more than 30 chickens.
The simple way to compute is to start with the average egg-weight
for each chicken, though it is conceivable to do weighting by the
n or the standard error.
 
Is there homogeneity among the egg-weights, across chickens? - you
could construct an ANOVA between chickens, and that would make
further use of "within-chicken"  data.
 
If there are within-chicken live-weights, then you could have
within-chicken regressions, between live-weight and egg-weight.
If these are not heterogeneous, then it would be fair to
average them.
 
 
 
Rich Ulrich, biostatistician                wpilib+@pitt.edu
http://www.pitt.edu/~wpilib/index.html   Univ. of Pittsburgh