HI Mauricio and Ecolog,

It seems you have 17 data points, and each has two variables x and y.
Five of these points have data for both x and y, and 12 points have data
on one of them (say x). You want to know correlation between x and y. Is
this the problem you are addressing? The simplest way (and safe
probably) is to use the five points with both x and y data to calculate
the correlation; but the sample size is very small (5 only), which may
or may not give you good insight into your question of interest.

It seem you had done some data imputation given what you said. I am not
aware of whether you have other variables or information that you can
use to "estimate the missing ones". If not, I am not sure what you did
was reliable. If you build a regression based on five points, and then
estimate the missing values of the 12 points based on this regression,
then pool the five known values and 12 estimated values (say for y). Now
you got 17 "data points" (but 12 points have "pseudo" data) and used
them for correlation calculation. I doubt this is a good way because
there is a loop in this way.

Best,

Li

On 2011-06-21 10:53, Mauricio Carrasquilla wrote:
 Greetings,



 I am working on two data sets and want to correlate them but I am having
 some issues and would like some help from this list. The problem is that I
 have one variable that has 5 values and the other has 17 and I want to
 correlate them. I’ve done it in several different ways and come out with
 different correlation coefficient and significance. I’ve tried regression
 for the variable with only 5 values to estimate the missing ones, I’ve  also
 tried introducing the same value on the missing “values “until the new one
 (on a time scale) and finally have tried a direct correlation with different
 sample size.



 I’ve been reading about correlations, regressions and missing data, but have
 not yet come up with a conclusive and strong response. I really think this
 list will be of great help for my doubts.







 Thanks in advance,



 Mauricio Carrasquilla

 Marine Biologist

 Universidad Jorge Tadeo Lozano, Colombia, SA

 Masters in Natural Resources and Environment (C)

 Instituto Politécnico Nacional (MX)

   <mailto:mauricio.carrasqui...@gmail.com>   mauricio.carrasqui...@gmail.com




* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Li An (安力), PhD
Associate Professor
Department of Geography
San Diego State University
http://www-rohan.sdsu.edu/~lian/ (Personal website)
http://complexity.sdsu.edu/ (Group Website)

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to