Book contents
- Frontmatter
- Contents
- Preface
- Acknowledgments
- Acronyms and Abbreviations
- Part I RNAi HTS and Data Analysis
- Part II Methodological Development for Analyzing RNAi HTS Screens
- 7 Statistical Methods for Group Comparison
- 8 Statistical Methods for Assessing the Size of siRNA Effects
- References
- Index
- Plate section
8 - Statistical Methods for Assessing the Size of siRNA Effects
Published online by Cambridge University Press: 03 May 2011
- Frontmatter
- Contents
- Preface
- Acknowledgments
- Acronyms and Abbreviations
- Part I RNAi HTS and Data Analysis
- Part II Methodological Development for Analyzing RNAi HTS Screens
- 7 Statistical Methods for Group Comparison
- 8 Statistical Methods for Assessing the Size of siRNA Effects
- References
- Index
- Plate section
Summary
The size of an siRNA effect is represented by the magnitude of difference between the siRNA and a negative reference. Traditionally, mean difference (or, equivalently, average fold change in log scale), along with p-value of testing mean difference, has been used to indicate siRNA effects. However, as a statistical parameter, mean difference does not contain any information about data variability, cannot effectively measure the magnitude of difference between two groups, and thus cannot be used to assess siRNA effects successfully. Recently, SSMD and d+-probability have been proposed for the comparison of two groups and have been extended to multigroup comparisons. SSMD is a special case of SMCV when only two groups are involved in a comparison. Thus, given the concepts and theorems regarding contrast variable, SMCV, and c+-probability presented in Chapter 7, I explore the use of SSMD and d+-probability for assessing the size of siRNA effects in this chapter. Specifically, I first present the concepts of SSMD and d+-probability along with their relationship in Section 8.1 and the estimation of SSMD in Section 8.2. Standardized mean difference has been used for measuring the magnitude of difference, and classical t-statistic has been used for selecting hits in screens with replicates. Both look similar to SSMD. Therefore, I compare SSMD with standardized mean difference and classical t-statistic in Section 8.3.
- Type
- Chapter
- Information
- Optimal High-Throughput ScreeningPractical Experimental Design and Data Analysis for Genome-Scale RNAi Research, pp. 154 - 188Publisher: Cambridge University PressPrint publication year: 2011