Stats 300 C : Theory of
نویسنده
چکیده
We have seen that the BH procedure controls the FDR for multiple testing when the p-values are independent. How can we control the FDR under dependence? Here, we consider the case where the statistics have distribution Y ∼ N (μ,Σ); in the past, we have assumed that Σ = Ip. If the covariance matrix Σ were known, we could reduce to the independent case using the transformation Y ′ = Σ− 1 2Y , so that
منابع مشابه
Stats 300 C : Theory of Statistics Spring 2017 Lecture 25 — Jun 5 , 2017
In today’s class we will consider the above problem (1) when μ is sparse. We begin by motivating the special case of sparsity. Thereafter we show that a thraueshold estimator is optimal (in minimax sense), if the sparsity of μ is known. We then introduce FDR-threshold estimator, which is oblivious to μ’s sparsity and is yet optimal. We then draw parallel between this problem and multiplehypothe...
متن کاملMODELLING AND ANALYSIS OF A DISCRETE-TIME PRIORITY QUEUING COMPUTER NETWORK WITH PRIORITY JUMPS USING PROBABILITY GENERATING FUNCTIONS
Priority queues have a great importance in the study of computer communication networks in which different types of traffic require different quality of service standards. The discrete-time non-preemptive priority queuing model with priority jumps is proposed in this paper. On the basis of probability generating functions mean system contents and mean queuing delay characteristics are obtained....
متن کامل