Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
1153180 | Statistics & Probability Letters | 2013 | 4 Pages |
Abstract
If XX and YY are independent and if X+YX+Y and X/(X+Y)X/(X+Y) are independent random variables, then XX and YY must have gamma distributions. To confirm that lack of correlation between XX and X/(X+Y)X/(X+Y) does not characterize the gamma distribution, a large class of distributions are identified for which cov[X,X/(X+Y)]=0. A related question in the context of matrix-variate distributions is addressed.
Related Topics
Physical Sciences and Engineering
Mathematics
Statistics and Probability
Authors
Barry C. Arnold, Jose A. Villasenor,