Robustness Analysis & Statistical Inference

boubaBouba Housseini Suman photoSuman Seth
  • Stochastic dominance for the Alkire Foster method
  • Rank robustness; Kendall’s Tau and Spearman
  • Rank concordance methods.
  • Computation of standard errors and confidence intervals

Audio 

Video (with guide)

Guide to the video

Introduction

Part 1 Robustness analysis

1:04 Sources for the lecture

3:28 Policy areas requiring robustness analysis

4:34 Importance of robustness analyses illustrated using Global MPI data

6:15 Implications of conclusions based on a sample

8:04 Parameters of M0 for robustness analysis: poverty cutoff, weighting vector and deprivation cutoffs

9:48 Rank robustness analysis

11:03 First Order unidimensional dominance

13:15 Dominance for H and M0 in AF

16:24 Calculation using two deprivation count vectors

21:10 Discussion of the M0 curve

23:24 Stochastic dominance conditions

26:42 Methods for comparing robustness of ranking

28:38 Kendall’s Tau

32:43 Spearman’s Rho

34:00 Some illustrations using the MPI: Robustness to weights

37:18 Robustness to poverty cutoff (k)

Part 2 Statistical inference

46:30 Estimation from samples about population level characteristics

47:18 Common concerns

49:20 Standard error (SE) and construction of confidence intervals (CI)

54:40 How to obtain the standard error

58:06 How to use a confidence interval

1:00:00 Difference between analytical SE and bootstrap methods

1:08:10 Concerns with sampling

1:11:38 Example using analytical method for SE using data from India

1:13:55 Hypothesis tests and equivalent procedures

1:19:50 Statistical inference in MPI comparisons

1:34:35 Conclusions