https://en.wikipedia.org/wiki/Variance-stabilizing_transformation
The variance stabilizing transform of a Poisson random variable is the Anscombe transform. What is the VST for a negative binomial random variable?
$\begingroup$
$\endgroup$
4
-
$\begingroup$ Re "The:" a variance stabilizing transform is not unique. For instance, the square root is another variance-stabilizing transform of the Poisson family. A general method for finding a variance-stabilizing transform is given at stats.stackexchange.com/questions/251449. $\endgroup$– whuber ♦Apr 14, 2022 at 21:50
-
$\begingroup$ I am looking for the best vst. $\endgroup$– user1424739Apr 14, 2022 at 22:00
-
1$\begingroup$ "Best" in what sense? $\endgroup$– whuber ♦Apr 14, 2022 at 22:01
-
$\begingroup$ In the sense that the Anscombe transform is better than the sqrt transform for Poison distribution. $\endgroup$– user1424739Apr 15, 2022 at 3:33
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Take a look at Anscombe's paper from 1948 the transformation of Poisson, Binomial and Negative-Binomial data. There, he shows that for a random variable $$ Y \sim \text{NegBin}(m, k), $$ where $m$ is the mean and $k$ is the size, the transformation $$ Z = \sinh^{-1}\sqrt{\frac{Y + c}{k - 2c}} $$ with $c \approx 3/8$ is in some sense optimal.
