The convergence approach at the core of the consensus process is fundamental to what the IPC describes as its “multi- sectoral cooperation and technical consensus”, a means of involving the interpretations of a range of stakeholders, incorporating context, overcoming data limitations in quality or availability, and achieving acceptance and ownership of the final classifications. [...] Understanding the accuracy of the IPC and modeling and analyzing the consensus process is a topical and important area of research not only for the IPC itself but also for policymakers and users of the IPC, including researchers. [...] IPC and food security: Nature of the problem of analyzing accuracy What, in the context of the IPC, and in the broader context of food security measurement and analysis, is meant by ‘accuracy’? Ideally, we would compare an IPC classification of an IPC analysis area and time period to the lived experience of food insecurity and malnutrition experienced by all people living in that location at that. [...] Analysts walk through the IPC analysis worksheet to analyze and discuss evidence related to their sector(s) and the IPC protocols, aiming to achieve a strong understanding of the underlying data, and to reach agreement around both the classification of IPC analysis areas and the proportion of population in each phase classification. [...] In the findings that follow, we define the impact of consensus as the area between the average of the food security indicators and the observed IPC outcomes.
- Pages
- 93
- Published in
- Italy
Table of Contents
- ANALYZING THE INTERNAL CONSISTENCY OF IPC AFI AND AMN ANALYSES 1
- ANALYZING THE INTERNAL CONSISTENCY OF IPC AFI AND AMN ANALYSES 2
- CHAPTER 1 INTRODUCTION 14 4
- CHAPTER 2 NATURE OF THE PROBLEM AND METHODS 17 4
- CHAPTER 3 AFI DATA AND SAMPLES 27 4
- CHAPTER 4 AFI RESULTS 35 4
- CONTENTS 4
- CHAPTER 5 AMN OVERVIEW 56 5
- CHAPTER 6 DISCUSSION LIMITATIONS 65 5
- CHAPTER 7 RECOMMENDATIONS 70 5
- ANNEXES ANNEX 1 WHAT AFI DATA WERE WE ABLE TO INCLUDE AND WHY 74 5
- ANNEX 2 WHAT AMN DATA WERE WE ABLE TO INCLUDE AND WHY 78 5
- ANNEX 3 ADDITIONAL FINDINGS 81 5
- ANNEX 4 BIBLIOGRAPHY 94 5
- ACKNOWLEDGEMENTS 6
- AFI 7
- AMN 7
- API 7
- CARI 7
- CDT 7
- FCS 7
- FEWS NET 7
- FIES 7
- FSNMS 7
- FSI 7
- GAM 7
- GSU 7
- HDDS 7
- HFA 7
- HHS 7
- IPC 7
- IPC-API 7
- LCS 7
- MUAC 7
- OLS 7
- TWG 7
- WHZ 7
- ABBREVIATIONS AND ACRONYMS 7
- EXECUTIVE SUMMARY 8
- 1.1. Objective and approach the role of consensus 11
- 1. INTRODUCTION 11
- 1.2. Roadmap 13
- 2.1. IPC and food security Nature of the problem of analyzing accuracy 14
- 2.2. Evaluating metrics when there is no objective measured truth 14
- 2. NATURE OF THE PROBLEM AND METHODS 14
- 2.2.1 Qualitative findings 15
- 2.2.2 Defining study 17
- 2.2.3 Study parameters 18
- 2.2.4 Note on language 18
- 2.3. Conceptual framework to understanding consensus 18
- 2.3.1. A simple model of the consensus process 20
- 2.4. Application of methods for current status AFI 20
- 2.4.1 Arithmetic simple mean 20
- 2.4.2 Estimating weights 21
- 2.4.3. Distributional analysis bunching 22
- 2.4.4 Residual analysis 22
- 2.4.5 Noisy input data 22
- 2.5 Methods for projected AFI 23
- 3.1. Data collection 24
- 3.2 Descriptive statistics current status AFI 24
- 3.2.1. Sample size FSI availability and FSI concordance 24
- 3. AFI DATA AND SAMPLES 24
- 3.2.2. Reliability score 28
- 3.2.3. Consensus-based phase classification outcome compared to FSI implied phase classification outcome 28
- 4.1 Results for current status AFI 32
- 4. AFI RESULTS 32
- 4.1.1 Arithmetic simple mean classification estimates compared to consensus-based classifications 33
- 4.1.2 Arithmetic simple mean population estimates compared to consensus-based population 35
- 4.1.3 Logit weights across data availability and severity 38
- 4.1.4 OLS Weights on population estimates 40
- 4.1.5 OLS weights across TWG analysis rounds within countries 41
- 4.1.6 Distributional analysis bunching 43
- 4.1.7 Residual analysis are there favored analysis areas with non-random influence. 47
- 4.1.8 Noisy Input Data 49
- 4.2 AFI projections 51
- 5. AMN OVERVIEW 53
- 5.1 AMN proof of concept data and methods 54
- 5.2 Descriptive statistics and level of evidence 55
- 5.3 Consistency between representative and non-representative survey based AMN phase classifications 56
- 5.4 Consistency of results among WHZ MUAC and AMN consensus-based classifications 58
- 5.5 Current status and projections AMN 60
- 6.1 AFI current status 62
- 6. DISCUSSION LIMITATIONS 62
- 6.2 AFI projections 64
- 6.4 Limitations 65
- 7.1 Short-term recommendations Ways to support accuracy and perception of accuracy 67
- 7. RECOMMENDATIONS 67
- 7.2 Long-term recommendations Undertake an accuracy study in five years 69
- 1.1 AFI outcomes from the IPC-API 71
- 1.2. Sampling Procedure 1 Evaluating the Convergence Process 71
- ANNEX 1 What AFI data were we able to include and why 71
- 1.3. Sampling Procedure 2 Identifying Distributional Bunching Patterns and Projection t-1 and Realized Current Status t Outcome Comparison 73
- 1.4. Reliability Score for AFI 73
- 2.1. Data Collection and Limitation 75
- ANNEX 2 What AMN data were we able to include and why 75
- 2.2 Data Sampling 77
- 3.1. FSI vs. Consensus-based 3 Population - country specific results 78
- ANNEX 3 Additional Findings 78
- 3.2. FSI vs. Consensus-based 3 Population - cases with 4 FSI 79
- 3.3 FSI vs. Consensus-based 4 Population - Sample 5 FSI 80
- 3.4. Difference between IPC 3 population and the arithmetic mean of the FSI implied 3 population - countryTWG specific results 81
- 3.5 Correlation of residuals to underlying characteristics of the input data 82
- 3.6. AFI Realized Current Status Analysis Classifications RCSA at time t vs. Projected Status PS from the prior quarter t-1 84
- 3.7. Distributional analysis Country specific bunching results 85
- 3.8 Additional results 87
- ANNEX 4 Bibliography 91
