sample size (n) for mark-recapture analyses

[Robertal]

A reviewer in a recent journal asked for statistical tests to evalaute significant differences in survival across groups. I was planning on using a simple Wald Stat, but am confused as to what I would use as my 'n' for a z-stat.

If the Wald stat is inappropriate, any suggestions would be great.
Thanks.
R

[cschwarz@stat.sfu.ca]

The Wald test simply looks at the difference in estimates using the se to compute the test statistic. The se implicitly includes the "sample size" so an extra "n" is not needed.

Let s1, s2 be the two estimated survival probabilities and se() their se.
You will need s1, s2, se(s1), se(s2), and covariance between s1 and s2.

THe Wald test simply computes
z = (s1-s2)/sqrt(se(s1)**2 + se(s2)**2 - 2cov(s1,s2))


HOWEVER, if you are interested in difference in survival rates among groups, this is just the test in mark for the group effect in survival rates. So the WALD test may not be needed.

HOWEVER, do you really need hypothesis testing? The AIC weights the models by the weight of evidence for each model (in the model set), so accounts for the amount of information about differences in survival.

Please read Chapter 6 of GIM.

YOu didn't give details about the experimental design in your post -- it is always dangerous to do "testing" without consideration of the experimental design.