Article
Summary:
First, under a multivariate normal distribution with all correlations of the form $Q_{ij}=b_ib_j$ (where $-1\leq b_i, b_j\leq 1$), the probabilities of certain convex symmetric regions are shown to be, roughly speaking, non-decreasing functions of $\left|Q_{ij}\right|$. Second, under an equicorrelated normal distribution, the probabilieties of certain regions (which need be neither convex nor symmetric) are shown to be non-decreasing functions of the correlations. Third, some inequalities for special cases of multivariate exponential and Poisson distributions are given.
References:
[5] C. G. Khatri:
Further contributions to some inequalities for normal distributions and their applications to simultaneous confidence bounds. Ann. Inst. Statist. Math. 22 (1970), 451 - 458.
DOI 10.1007/BF02506363 |
MR 0283913 |
Zbl 0294.62013
[7] Z. Šidák: Unequal numbers of observations in comparing several treatments with one control. (In Czech.) Apl. Mat. 7 (1962), 292-314.
[8] Z. Šidák:
On probabilities of rectangles in multivariate Student distributions: their dependence on correlations. Ann. Math. Statist. 42 (1971), 169-175.
DOI 10.1214/aoms/1177693504 |
MR 0278354
[9] Z. Šidák:
A note on C. G. Khatri's and A. Scott's papers on multivariate normal distributions. Submitted to Ann. Inst. Statist. Math.
Zbl 0368.62029
[10] Z. Šidák:
A chain of inequalities for some types of multivariate distributions, with nine special cases. Apl. Mat. 18 (1973), 110-118.
MR 0315842