Dataset Information

A simple and robust method for partially matched samples using the p-values pooling approach.

ABSTRACT: This paper focuses on statistical analyses in scenarios where some samples from the matched pairs design are missing, resulting in partially matched samples. Motivated by the idea of meta-analysis, we recast the partially matched samples as coming from two experimental designs and propose a simple yet robust approach based on the weighted Z-test to integrate the p-values computed from these two designs. We show that the proposed approach achieves better operating characteristics in simulations and a case study, compared with existing methods for partially matched samples.

SUBMITTER: Kuan PF

PROVIDER: S-EPMC3717400 | biostudies-literature | 2013 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A simple and robust method for partially matched samples using the p-values pooling approach.

Kuan Pei Fen PF Huang Bo B

Statistics in medicine 20130217 19

This paper focuses on statistical analyses in scenarios where some samples from the matched pairs design are missing, resulting in partially matched samples. Motivated by the idea of meta-analysis, we recast the partially matched samples as coming from two experimental designs and propose a simple yet robust approach based on the weighted Z-test to integrate the p-values computed from these two designs. We show that the proposed approach achieves better operating characteristics in simulations a ...[more]

PMID: 23417968

Similar Datasets

Project description:Background: As costs decline, the size and scope of microarray experiments have increased. In multi-centre studies there is a need to ensure consistency of data pre-processing across centres. Similarly, in smaller scale studies the evolution of microarray platforms means that there is often a need to compare data generated on earlier microarrays to that generated on newer ones. It is important in such studies to ensure that platform-dependent biases are removed so that meta-analysis of different datasets can be performed reliably. In both these cases the optimal scenario is to have a small subset of samples repeated at each site or on each platform. These replicates can then be used to learn a relationship between probe intensities on the two platforms. Results: I introduce here a simple, linear-modelling-based method for normalizing data from multiple-platforms by using replicate hybridizations. A dataset of 20 rat liver samples is used as a benchmark. Eight samples are hybridized to two separate versions of Affymetrix microarrays, while the other 12 are hybridized to one, for a total of 28 arrays. Our linear modelling method removes platform bias as assessed using both unsupervised machine-learning and two-group statistical analyses. The method is computationally efficient and works well for data pre-processed by the GCRMA, RMA and MAS5 algorithms and using either default or alternative probe-mappings. The method is very stable towards the number of replicate samples used, with even two replicates greatly reducing platform-specific bias. Conclusions: A simple linear-modelling method can remove platform-specific bias independent of the pre-processing algorithm and ProbeSet-mapping used. This technique can readily be extended to multi-site experiments, and suggests the benefits of including a small number of replicate hybridizations in each new study as a normalization control. Twenty rats livers were processed, eight on both RAE230-A and RAE230-2 arrays, 8 on only RAE230-A arrays, and 4 on RAE230-2 arrays only.

Dataset Information

A simple and robust method for partially matched samples using the p-values pooling approach.

Publications

A simple and robust method for partially matched samples using the p-values pooling approach.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure