Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
"Robert Fornango" <email@example.com>
Thu, 13 Dec 2012 09:21:32 -0700
"I am having is that when I correlate these at each time point seperately
ie. 'correlate var0a var0b' where var0a is toxinA serum level at baseline
and var0b is toxinB serum level at baseline, there is no correlation.
Therefore is it right to assume that the association is either weak (ie.
requires all three time points for more data) or not linear...?"
You are correct. Not having seen any of the results, I would venture to say
that "no correlation" between the individual time points for Toxin A and
Toxin B means either a) the association is very weak over time (if there is
one at all), or b) it is a non-linear association.
Not to be mean, but I'm surprised you got to the point of running -xtmixed-
and -iccvar- without establishing this first.
If you want to obtain Pearson's r for the pooled data, you can -reshape-
your data into long format (the same used for -xtmixed- with longitudinal
data), and use -correlate-.
As you noted in the quote above, if the association is non-linear, then one
of the assumptions of Pearson's r is violated and that statistic is not
appropriate. However, if the relationship is non-linear, I would think you
would detect something in the cross-sectional correlations (i.e. between
var0a var0b, var1a var1b, etc.).
It might be useful to obtain a line plot of the relationship between toxin
levels over time. For example, if you have the data in long form, with a
single column for toxin A (var_a) and another for toxin B (var_b), and "id"
is your patient identifier, then you could try:
twoway (line var_a var_b)
If you want to examine the plot for each patient individually, try:
twoway (line var_a var_b), by(id)
If there truly is no relationship between the toxin levels over time, your
plots should not display any real pattern (e.g. the proverbial "mess").
I hope this helps.
Robert J Fornango, PhD
Chief Executive Officer
F1 Analytics LLC
* For searches and help try: