Atnaujinkite slapukų nuostatas

Statistical Survey Design and Evaluating Impact [Kietas viršelis]

, ,
  • Formatas: Hardback, 214 pages, aukštis x plotis x storis: 251x189x20 mm, weight: 660 g
  • Išleidimo metai: 09-May-2016
  • Leidėjas: Cambridge University Press
  • ISBN-10: 1107146453
  • ISBN-13: 9781107146457
Kitos knygos pagal šią temą:
Statistical Survey Design and Evaluating ImpactDidesnis paveikslėlis
  • Formatas: Hardback, 214 pages, aukštis x plotis x storis: 251x189x20 mm, weight: 660 g
  • Išleidimo metai: 09-May-2016
  • Leidėjas: Cambridge University Press
  • ISBN-10: 1107146453
  • ISBN-13: 9781107146457
Kitos knygos pagal šią temą:
Pastovi nuoroda: https://www.kriso.lt/db/9781107146457.html
Raktažodžiai:
Statistical designs, sample surveys and evaluation designs are fundamental tools for solving queries related to population parameters and the effects of public programs and policies. This book explores the concepts of effective sampling and evaluation techniques in a cohesive and concise manner. Sampling design techniques, including simple random sampling, stratified sampling, systematic sampling and cluster sampling, are presented in detail. These techniques play a vital role when choosing an appropriate sample survey design. The concepts of multistage design, non-sampling errors and evaluation techniques including before-after design, one-time treatment and control design are discussed extensively. The book focuses on different methods of estimation, including multiple regression analysis and logistic regression. It covers the issue of bias in a design, the source of such bias and ways to overcome it. Clear guidelines with remedial measures are outlined to facilitate choosing a suitable sampling design.

Daugiau informacijos

This book discusses important methodologies for developing statistical designs, sample surveys and evaluation designs.
Figures
xv
Tables
xvi
Foreword xvii
Preface xix
Acknowledgments xxiii
1 Introduction to Sample Survey Designs
1.1 Introduction
1(1)
1.2 Population, Units and Sampling Units
1(2)
1.3 Sampling Design
3(1)
1.4 Probability and Purposive Sampling
3(2)
1.4.1 Probability sampling
3(1)
1.4.2 Purposive sampling
4(1)
1.5 Frame
5(1)
1.6 Bias and Error
6(5)
1.7 Few Guidelines for a Desirable Sampling Design
11(2)
2 Basic Sampling Designs
2.1 Introduction
13(1)
2.2 Simple Random Sampling
14(17)
2.2.1 Description
14(1)
2.2.2 Methods of selection
15(1)
2.2.3 Estimation of mean, total and need for weights
16(3)
2.2.3.1 Normalization of weights
19(2)
2.2.3.2 Role of weights
21(1)
2.2.4 Estimation of proportion
22(1)
2.2.5 Subclass estimates
22(1)
2.2.6 Sampling variance of estimates
23(1)
2.2.6.1 Sampling variance of a sample mean
23(2)
2.2.6.2 Sampling variance of estimated population total
25(1)
2.2.6.3 Sampling variance of proportion
25(1)
2.2.6.4 Sampling variance of subclass estimates
25(3)
2.2.7 Determination of sample size
28(3)
2.3 Stratified Sampling M
2.3.1 Description
31(2)
2.3.2 Estimation of parameters
33(1)
2.3.2.1 Estimation of mean
34(1)
2.3.2.2 Estimation of total
35(1)
2.3.2.3 Estimation of proportion
35(1)
2.3.3 Weighting and its similarity with standardization
35(2)
2.3.4 Sampling variance of estimates
37(1)
2.3.4.1 Sampling variance of mean
38(1)
2.3.4.2 Sampling variance of total
38(1)
2.3.4.3 Sampling variance of proportion
38(1)
2.3.5 Allocation and selection of units
38(1)
2.3.5.1 Proportional allocation
39(1)
2.3.5.2 Optimum allocation
40(1)
2.3.5.3 Practical guidelines for allocation
41(1)
2.3.6 Some advantages of stratification
42(1)
2.3.7 Post-stratification
43(2)
2.4 Systematic Sampling
45(5)
2.4.1 Description
45(1)
2.4.2 Method of selection
45(1)
2.4.2.1 Decimal interval method
46(1)
2.4.3 Advantages of systematic sampling
46(1)
2.4.4 Disadvantages of systematic sampling
47(1)
2.4.4.1 Monotonic trend
47(1)
2.4.4.2 Periodicity
47(1)
2.4.5 Estimation of parameters and their sampling variances
48(1)
2.4.5.1 Two consecutive units per stratum
49(1)
2.5 Probability Proportional to Size Sampling
50(4)
2.5.1 Description
50(4)
2.6 Cluster Sampling
54(6)
2.6.1 Description
54(1)
2.6.1.1 Preparation of artificial clusters
55(1)
2.6.2 Method of selection
56(1)
2.6.3 Estimation of parameters and sampling variances
57(1)
2.6.3.1 Clusters of equal size
57(1)
2.6.3.2 Clusters of unequal size
58(2)
2.7 Key Points
60(2)
3 Multi-stage Designs
3.1 Introduction
62(1)
3.2 Two-stage Design with Equal Size Clusters
63(16)
3.2.1 Components of overall variation
64(5)
3.2.2 Two-stage design for selection of units with equal probability
69(1)
3.2.2.2 Estimation of mean and variance
70(3)
3.2.2.3 Clustering, design effect and choice of number of PSUs/cluster
73(6)
3.3 Two-Stage Design with Unequal Cluster Size
79(17)
3.3.1 Estimation of mean and sampling variance
80(4)
3.3.2 Two desirable properties of the design
84(1)
3.3.3 Guidelines for attaining desired property
85(2)
3.3.4 Ways to control variations in cluster size
87(1)
3.3.4.1 Controlling size of clusters
87(1)
3.3.4.2 Alternative selection procedure
87(2)
3.3.4.2.1 Alternatives when information on size of PSUs refers to a past period
89(7)
3.3.4.3 Stratification of PSUs to reduce variations in cluster size
96(1)
3.4 Stratification in Multistage Design
96(7)
3.4.1 Estimation of parameters in unequal cluster size
97(3)
3.4.2 Estimation of parameters in equal cluster size
100(3)
3.5 Selection of Sampling Units at Different Stages
103(4)
3.5.1 Selection of PSUs
103(1)
3.5.2 Selection of second-stage units
104(1)
3.5.3 Selection of individuals within a household
105(2)
3.6 Key Points
107(2)
4 Probability Sampling under Imperfect Frame
4.1 Introduction
109(1)
4.2 Sampling Populations Having Specific Attributes
110(7)
4.2.1 Sampling when target population is not rare
111(1)
4.2.1.1 Sampling without screening
111(1)
4.2.1.2 Sampling with screening
112(1)
4.2.1.3 Relative advantages with screening and without screening
112(1)
4.2.1.4 Facilitating screening
113(1)
4.2.2 Sampling for rare attributes
114(1)
4.2.2.1 Household-based sampling of rare population
114(3)
4.3 Defective Frame
117(5)
4.3.1 Duplications
118(1)
4.3.1.1 Estimation in presence of duplications
118(1)
4.3.1.2 Procedure to deal with duplicate listing
119(2)
4.3.1.3 Incompleteness or omissions in a frame
121(1)
4.4 Sampling in Absence of a Frame
122(6)
4.4.1 Facilitating a cluster design
123(2)
4.4.2 Selection, data collection and estimation
125(3)
4.5 Household Listing
128(3)
4.5.1 Two alternatives if listing is to be avoided
129(2)
4.6 Key Points
131(2)
5 Tackling Non-Sampling Errors
5.1 Introduction
133(1)
5.2 Coverage Error
134(3)
5.2.1 Discrepancy between study and target population
134(1)
5.2.2 Omission of areas to reduce cost
135(1)
5.2.3 Tackling small PSUs
135(1)
5.2.4 Error in identification of a PSU
136(1)
5.2.5 Error in segmentation of a PSU
136(1)
5.2.6 Error in listing a PSU/segment
137(1)
5.3 Non-response Error
137(6)
5.3.1 Remedy for non-response
139(1)
5.3.1.1 Adjustment for non-response when error is randomly distributed
140(1)
5.3.1.2 Adjustment for non-response when error is not completely random
141(1)
5.3.2 Item non-response error
142(1)
5.4 Response Error
143(5)
5.4.1 Questionnaire construction
143(1)
5.4.1.1 Factual questions
143(1)
5.4.1.2 Non-factual questions
144(2)
5.4.2 Errors due to investigators
146(1)
5.4.2.1 Training
147(1)
5.4.2.2 Supervision
147(1)
5.5 Key Points
148(1)
6 Introduction to Evaluation Design
6.1 Background
149(1)
6.2 Bias and Error
150(5)
6.2.1 Bias
150(3)
6.2.2 Bias elimination
153(2)
6.2.3 Error
155(1)
6.3 Types of Evaluation Designs
155(3)
6.3.1 Evaluation designs with random allocation of units
155(1)
6.3.2 Evaluation designs with clusters allocated randomly
156(1)
6.3.3 Evaluation designs with unit level matching
156(1)
6.3.4 Evaluation designs with cluster matching
157(1)
6.3.5 Observational and case-control studies
157(1)
7 Designs for Causal Effects: Setting Comparison Groups
7.1 Introduction
158(1)
7.2 Measuring Main and Interaction Effects
159(4)
7.3 Bias and Error in Measurement of Treatment Effect
163(5)
7.3.1 Sources of bias
164(3)
7.3.2 Internal and external validity
167(1)
7.4 Three Basic Designs for Estimating Treatment Effect
168(11)
7.4.1 One sample each from T and C at two different times (before-after design)
168(200)
7.4.1.1 Description and estimation of effect
368
7.4.1.2 Biasing effects and remedies
169(4)
7.4.1.3 Estimation of standard error of estimated impact
173(1)
7.4.2 One sample each from T and C observed at one point of time (treatment-control design)
174(1)
7.4.2.1 Description and estimation of effect
174(1)
7.4.2.2 Biasing effects and remedies
175(1)
7.4.2.3 Estimation of standard error of estimated impact
175(1)
7.4.3 Two samples each from T and C observed at two points in time (before-after and treatment-control design)
175(1)
7.4.3.1 Description and estimation of effect
175(2)
7.4.3.2 Biasing effects and remedies
177(1)
7.4.3.3 Estimation of standard error of estimated impact
178(1)
7.5 Output and Its Timing
179(1)
7.6 Key Points
179(2)
8 Designs for Causal Effects: Allocation of Study Units
8.1 Introduction
181(1)
8.2 Alternative Tools to Attain Balance
181(3)
8.2.1 Randomization
182(1)
8.2.2 Stratification
183(1)
8.2.3 Pair matching
184(1)
8.3 Advantages and Disadvantages of Three Tools
184(3)
8.3.1 Randomization
185(1)
8.3.2 Matching
185(1)
8.3.2.1 Stratification
185(1)
8.3.2.2 Pair matching
186(1)
8.4 Choice of Study Units
187(4)
8.4.1 Procedure of allocation of units/clusters
188(1)
8.4.1.1 Randomization
188(1)
8.4.1.1.1 Restricted randomization
189(2)
8.4.1.2 Stratification
191(1)
8.4.1.3 Pair matching
191(1)
8.5 Potential Outcome Framework
191(5)
8.5.1 Propensity score matching
192(4)
8.6 Choice of a Design
196(2)
8.7 Key Points
198(1)
9 Statistical Tests for Measuring Impact
9.1 Introduction
199(2)
9.1.1 Two different ways to estimate impact
200(1)
9.2 Impact when Units are Allocated Randomly
201(15)
9.2.1 Testing difference between two means
202(1)
9.2.1.1 Testing means from two different populations
203(1)
9.2.1.2 Large-sample z-test
204(1)
9.2.1.3 Testing several means Application of ANOVA
205(2)
9.2.1.4 Non-parametric tests
207(6)
9.2.2 Testing difference between two proportions
213(1)
9.2.2.1 Chi-square test of independence
213(2)
9.2.2.2 Testing odds ratio
215(1)
9.3 Impact when Clusters are Allocated Randomly
216(2)
9.3.1 Analysis at cluster level
216(1)
9.3.2 Analysis at individual level
217(1)
9.4 Impact when Stratification is used before allocation
218(4)
9.4.1 When units are allocated
218(1)
9.4.1.1 Two-way ANOVA test
219(3)
9.4.2 When clusters are allocated
222(1)
9.5 Impact in Pair Matching
222(5)
9.5.1 Variables measured in interval scale
223(1)
9.5.2 Dichotomous variable
223(1)
9.5.2.1 Exact binomial test
224(2)
9.5.3 Non-parametric test
226(1)
9.6 Model-Based Analysis
227(6)
9.6.1 Multiple regression analysis
228(1)
9.6.1.1 Modifications in the case of cluster sampling
229(1)
9.6.2 Logistic regression
230(2)
9.6.3 Assumptions in regressions
232(1)
9.7 Key Points
233(1)
10 Case Studies
10.1 Introduction
234(1)
Part I Sample Survey Designs
234(1)
10.2 National Family Health Surveys, India (NFHS, India)
234(1)
10.3 Sampling Design of NFHS
234(13)
10.3.1 Sample size
234(2)
10.3.2 Choice of PSU
236(1)
10.3.3 Design for rural area
236(1)
10.3.3.1 Merging of small villages
236(1)
10.3.3.2 Stratification
237(2)
10.3.3.3 Selection of sampling units
239(4)
10.3.3.4 An alternative two-stage selection
243(1)
10.3.4 Design for urban area
243(1)
10.3.5 Estimation
244(1)
10.3.5.1 Computation of weights
244(1)
10.3.5.2 Estimation of parameters
245(2)
10.4 Other Global Large-Scale Surveys
247(2)
10.4.1 Use of master sample in survey designs
247(2)
10.4.2 Example of GATS sample design in Nigeria
249(1)
10.5 National Sample Surveys (NSS) in India
249(1)
Part II Evaluation Design
250(1)
10.6 Illustration of Evaluation Designs
250(5)
10.6.1 Impact evaluation of a life skills education intervention on adolescent girls' empowerment
250(2)
10.6.2 Fisher's tea test
252(1)
10.6.3 Impact of an intervention to promote use of Intra Uterine Device in a population
253(1)
10.6.4 An experiment to test effectiveness of a medicine
253(1)
10.6.5 An intervention to reduce post-partum haemorrhage
254(1)
References 255(5)
Index 260
Tarun Kumar Roy is former Director of the International Institute for Population Sciences (IIPS) in Mumbai. He served as a faculty member at IIPS for more than three decades and taught courses on sampling design and the evaluation of family welfare programs to graduate students. He was actively engaged in conducting several large-scale surveys, and was engaged in the National Family Health Survey (also known as the India Demographic and Health Survey). He serves as an expert on the Sampling Review Committee for the Global Adult Tobacco Survey (GATS), being conducted by the National Foundation for the Centers for Disease Control and Prevention in Atlanta, USA. Rajib Acharya is a statistician and demographer with more than fifteen years' experience in academic research, and planning, designing, managing and monitoring large studies. He worked as a Research Advisor at Johns Hopkins University from 2002 to 2004. He has expertise in the design of large sample surveys, sample size determination, processing and managing large-scale cross-sectional and longitudinal data sets and the analysis of survey data on population, health and nutrition. He has published a number of articles in national and international journals. Arun Roy is Chief Executive of Economic Information Technology, a socio-economic research organization in Kolkata. He received an MSc in Statistics from Patna University, an MA in Economics from Delhi University, and a PhD in Sociology from Kalyani University. He has worked as an economist at the National Council of Applied Economic Research, New Delhi, and has served as a Monitoring and Evaluation Consultant to several organizations, including the World Bank in Odisha, the Institutional Council on Management of Population Programs in Malaysia, and German Agro Action in Bonn.