1. FISHER'S Z: Using the formula for Fisher's z transformation, I am
able to estimate the limits for the expected values nicely. However, I
am unable to transform these back to limits for the actual interval for
the estimate. This part of the function would need to perform the same
task as Table B.8 in Neter, Kutner, et al. 4th ed. (Applied Linear Stat.
Models) where one transforms from z-prime to r as described on page
643. Applying the (rearranged) Fisher's z formula for this is not
working for negative values of the correlation statistic because of the
logs and *I was hoping someone someone might be willing to help me with
the math* (I'm fairly sure this is a mathematical problem). *Or is
there an S-Plus function somewhere that will do this?* I would like
this to be done within the function, although of course I could do the
transformation using the abovementioned table.
2. BOOTSTRAPPING: Patrick Burns suggested "bootstrapping the statistic."
I don't know much about it and I have two questions about this approach
--*First, would bootstrapping Pearson's result in a robust estimate similar to Spearman's, or should bootstrap be applied to Spearman's? (cor will not produce Spearman's; bootstrap will not work on cor.test because it does not give a vector - I can work this out if necessary).*
*Second, and most important, I am having problems deciding how I can use the bootstrap results to calculate intervals -- I have not yet got the references mentioned in the Splus manual, but I am the impatient type.*
============================================== Steve Bousquin
Colorado State University Department of Rangeland Ecosystem Science/GDPE NR209 Fort Collins, CO 80523 USA
sbous@lamar.colostate.edu ==============================================
----------------------------------------------------------------------- This message was distributed by s-news@wubios.wustl.edu. To unsubscribe send e-mail to s-news-request@wubios.wustl.edu with the BODY of the message: unsubscribe s-news