M. Nurul Islam An Introduction to SAMPLING METHODS Theory and Applications ae AVI ie ate@ iene CoeABOUT THE AUTHOR Dr. M. Nurul Islam is former Selection Grade Professor of Statistics, Faculty of Science at the University of Dhaka, Bangladesh. Currently Dr. Islam is Pro-Vice Chancellor at the World University of Bangladesh. He joined the University of Dhaka as a Lecturer in 1971 and acted as Chairman of the Department of Statistics during 1985-1987. He held the coveted position of the Vice Chancellor of the Mawlana Bhashani Science and Technology University in Tangail during 2009-2013. He is an elected member of famed International Statistical Institute (ISI) of the Netherlands. Dr. Islam is now holding the position of President of the Bangladesh Statistical Association. He also held the same poison of the Association during 2010-2012. Dr. Islam was an elected member of the Dhaka University Senate for the period 2002-2004. He was also a part-time member of University Grants Commission during 2010- 2012. During his 45 years' career as a teacher and researcher, he has written nearly one hundred scientific papers and survey reports related to his discipline, which have been published in both local and international journals. He has supervised a number of M.Sc and PhD works. He provided consultancy services to USAID, DANIDA, UNICEF, CIDA, CARE, SIDA, NETZ, Bangladesh Bureau of Statistics (BBS), Ministry of Health and Family Welfare and Ministry of Textile. He independently conducted a number of field surveys as a Principal Investigator. He extensively toured a number of countries including USA, UK, Canada, Egypt, Indonesia, Singapore, Nepal and India and presented papers there in conferences and seminars. He is author of four textbooks entitled (i) An Introduction to Statistics and Probability, (ii) An Introduction to Research Methods, (iii) An Introduction to Sampling Methods and (IV) An Introduction to Demographic Techniques published by Mullick & Brothers. The author has also published his entire research works in a single volume entitled Demographic Research in Bangladesh: An Update, with the financial assistance from Social Science Research Council, Ministry of Planning. He has been an Executive Editor of several national and international peer reviewed journals. Dr. Islam is also the recipient of the UGC Award and the Dhaka University Faculty Award for 2007 and 2004 respectively for two of his incomparable textbooks.CONTENTS PREFACE INTRODUCTION Concept of Samplin: Importance of Sampling... . Sampling: Historical Perspectives... Technical Terms.. Sampling Frame... Census and Survey... . Steps in Planning and Executing a 3 Sample Survey Advantages of Sampling over Complete Count Limitations of Sampling... > Sample Design and Survey Design... Evaluation of a Sample Design... . mes Sampling with and without Replacement... ... Confidence Interval... Probability and Non-probal Sampling Distribution... SIMPLE RANDOM SAMPLING Introduction... 0.0 66. eee eee eee cee oe Properties of Simple Random ‘Sampling... Drawing a Simple Random Sample... Estimation of Population Mean and Total Properties of the Estimators... Relative Error of the Estimators... Improving the Estimates... Advantages and Disadvantages of SRS... MORE ON SIMPLE RANDOM SAMPLING Introduction... Preliminaries. ae Variance of the Estimators... . Estimating Sample Size in SRS... ... we Developing a Formula for Sample Size... ity Sample... Sample Size for Estimating Population Proportion... Sample Size for Continuous Data. Problems in Estimating of Probability Proportional to Size Sampling it in SRS vii3.10 3.11 44 42 43 44 45 46 47 48 49 4.10 5.1 5.2 5.3 5.4 55 5.7 58 5.9 5.10 5.11 5.12 5.13 6.1 6.2 6.3 64 6.5 6.6 6.7 6.8 Estimated Variance for PPS Samplin: Design Effect... STRATIFIED SAMPLING Concept of Stratification... Principles of Stratification... Steps Involved in Stratified sampling... Estimators and Their Properties... Confidence Interval... Estimating Sample Size... Allocating Sample Size to Strata. : Disproportionate Stratified Sampling. Precision of Stratified Sampling... s Advantages and Disadvantages of Stratifi ied Sampling... RATIO ESTIMATES Introduction... Ratio Estimation under Simple Random Sampling. , Variance of the Ratio Estimator... Sample Estimates of Variance. Confidence Interval... Measuring the Efficiency of Ratio Estimator Bias in Ratio Estimation... Ratio Estimation in Stratified Sampling... Ratio Estimation under Modified Random Sampling... & Unbiased Ratio Type Estimator Product Estimator.. Multivariate Ratio Estimator... Comparing the Difference of Two Estimators... REGRESSION ESTIMATES Linear Regression Estimator... ... ... Properties of Regression Estimators... ‘ Sampling Variance of Regression Estimator ... Comparing Regression Estimator with Ratio and Mean Per Unit Estimators .. Bias of the Regression Estimator... Regression Estimator in Stratified Sampling... Limitations of Separate and Combined Estimators Difference Estimator.. viii 96 102 108 108 110 112 114 1341 134 139 144 162 164 474 171 175 183 188 190 193 198 204 210 216 217 219 221 238 238 241 244 248 251 252 2616.9 71 Ts 7.3 74 7.5 76 77 78 7.9 7.10 7.11 7.12 713 7.14 8.1 8.2 8.4 8.5 8.6 8.7 8.9 8.10 8.11 8.12 8.13 8.14 8.15 8.16 8.17 8.18 8.19 8.20 8.21 Choice of Estimators... SYSTEMATIC SAMPLING Introduction .. How to Draw a 5 Systematic: Samilet.. s Systematic Sampling in Relation to Other Designs Advantages and Disadvantages of Syatematic Sampling... “ Applications of Systematic ‘Sampling. Sample Mean and Its Variance... Implications of py on V(Ysy)... --- Comparison with Stratified Sampling... ... 0... Relative Precision of Systematic Sample Mean... Estimating Population Characteristics... Estimated Variance .. 2 Systematic sampling in Special populations, Repeated Systematic Sampling... ... PPS Approach in Systematic Sampling .. SINGLE-STAGE CLUSTER SAMPLING Introduction.. q Some Important Features of ‘Cluster Sampli ing... Reasons for Using Cluster Sampling... Disadvantages of Cluster Sampling. Cluster Sampling in Relation to Other Designs .. Applications of Cluster Sampling... ... ... .- Clustering Principles... . Simple One-stage Cluster samplin Variance in Terms of intra-class Correlation. Design Effect and Intra-class Correlation. Relationship between p, Deff. and M... Efficiency of Cluster Sampling... Optimum Cluster Size... 8 Variable-Size Cluster Sampling... Cluster Sampling with PPS... Horvitz-Thomson Estimator... on Estimation of Proportions in Cluster Sampling... Estimating Sample Size for Cluster Sampling... Relative Efficiency of Unequal Cluster... Use of Stratification in Cluster Sampling... Systematic Sampling Approach in Cluster Sampling 289 290 292 295 300 305 312 315 319 319 321 322 322 323 324 324 326 334 337 337 338 345 349 357 364 377 384 387 388 3909.1 9.2 9.3 9.4 9.5 96 9.7 9.8 10 10.1 10.2 10.3 10.4 10.5 10.6 10.7 10.8 10.9 10.10 10.14 10.12 10.13 10.14 10.15 10.16 10.17 10.18 1019 10.20 10.21 10.22 10.23 10.23 10.24 TWO-STAGE SAMPLING Introduction... Drawing Two-stage Cluster Sampling... Advantages of Two-stage Sampling... Estimation of Population Mean and Total . A Simple Interpretation of Two-stage Sample... Estimating a Population Proportion... Estimates for Equal-Sized Clusters. Two-stage Sampling with PPS... FURTHER ASPECTS OF SAMPLING Introduction... ... & Non-probability Samplin, Multi-stage Sampling... Multi-phase Sampling... Post-stratification. .. Area Sampling... on Acceptance Sampling... ... Pre-testing... Pilot Survey. Design Weight... Errors in Sample Survey... . Estimation of Variance in Complex Designs Questionnaire and its Construction... Steps Involved in Designing a Questionnaire... Important Surveys in Bangladesh: An Overview Demographic Surveys... Contraceptive Prevalence Surveys CPS)... Bangladesh Demographic and Health Survey... Bangladesh Health and Demographic Survey ... Health and Nutrition Survey... : . Bangladesh Child Nutrition Survey... Bangladesh Maternal Mortality Surveys Labour Force Survey... Household Expenditure Sune Bootstrapping ... RANDOM NUMBER BIBLIOGRAPHY . SUBJECT INDEX 394 394 392 396 398 402 406 412 428 470 4am 474 478 482 482 485 485 487 488 491 498 499 501Chapter 1 INTRODUCTION 1.1 CONCEPT OF SAMPLING During the past 30 years or so, the methods and techniques of sampling have reached a high level of scientific development. As a result, the uses of sampling have been extended into a wide variety of fields. From the standpoint of statistical data collection, sampling is a means for selecting a relatively small number of households, persons or other units for inclusion in a survey of some kind and inferring conclusions on the basis of these limited number of instances. This selection is done because enumeration of all units in the target population (population for which information is needed) is a large and complex undertaking that is almost always affected by limitations of time, budget and availability of experienced personnel. Not only that, it is unnecessary as well from the standpoint of precision and statistical reliability. Many countries have found, moreover, that sampling can play an important role in an overall census program (UN, 1997). Let us now introduce the concept of sampling by an example. EXAMPLE 1.1: Very frequently we talk about banning or restricting students’ politics in the university campus. This is a very sensitive issue. We sometime wonder whether our views on this issue are shared by the student community, who are directly or indirectly involved in this important issue. We may want to know the actual percentage of students of Dhaka University who do not approve of banning students’ politics in the campus. This percentage could be obtained by asking every student in the campusSTATISTICS AND ITS ORIGIN 3 nadequate. The works of Gosset appeared in Biometrika in 1908 under the same of ‘Student’, Gosset's pseudonym. (A famous story has it that Gosset was afraid to publish under his own name for fear that his employers, the inness Brewery, would be unhappy to discover that one of its chemists was doing research in statistics.) Today, student's ¢ is a basic tool of statisticians and experimenters. R. A. Fisher (1890-1962), who is known as the father of statistics, made numerous and significant contributions to statistics. He was influenced by Karl Pearson and Gosset, the two world known statisticians. His cioneering works made valuable contributions to evolve statistical cedures in many fields particularly agriculture, biology and genetics. is known for his contribution in the field of Analysis of Variance and Experimental Design. Meanwhile, Francis Galton (1822-1921) gave the concept of regression line while working on heredity of men and laws zoverning the transmission of physical and mental characteristics from one eration to another’. His work was confirmed by his friend Karl Pearson. The development of the theory of statistics was also effected by the pioneering works of a great number of authors, among whom Pareto, Adam, Edgeworth, Bowley, Yule, Stuart and Kendall deserve special mention. J. Neyman (1894-1981) and E. S. Pearson (1895-1980) will remain ever known for their original contributions to the theory of testing statistical hypothesis in 1936 and 1938. This theory promoted considerable research works of practical use. Indian statistician Mahalanobis (1893-1972) contributed significantly to the field of sample survey. He is well known for his D* statistic. He established Indian Statistical Institute (ISI) in 1931 and founded Sankhya, an international journal in Statistics. Qazi Motahar Hussain (1897-1981), an authority in statistics, made significant contributions towards the development of statistics in Bangladesh. He is known worldwide for his works "Hussain Chain’. He was the founder of the Department of Statistics and the Institute of Statistical Research and Training at the University of Dhaka. Deming's (1900-1993) "philosophical thoughts towards improving the quality of manufactured products through statistical quality control techniques is noteworthy. Genichi Tagguci (1924-) promoted the use of experimental designs for product improvement.CHAPTER 1 STATISTICS AND ITS ORIGIN 1.1 STATISTICS: ITS ORIGIN AND DEVELOPMENT It is not precisely known how the word statistics was originated. However, most people believe that the term statistics, derived from word state, was used to refer to a collection of facts of interest to the state. Some believe that the word statistic has been originated from the Italian word statista, the French word statistique and the German word statistik. This background tends to suggest that the term statistics has its origin from the ancient time. At that time, the word statistics had been believed to be in use as an indicator or yardstick of a country’s economic, political and social conditions. In this sense, the word statistics serves as an index of a country’s overall condition. In modern days too, relevant statistical data are: of immense importance to understand the level of development of a country. Presumably, all cultures that recorded history internationally also recorded statistics. The term statistics is an old one. As people abandoned their nomadic way of life, and started to live in groups, they urgently felt the need to know each other’s wealth, manpower, strength and position for their survival and safety. The group chiefs or rulers used to collect information on the above aspects in order to impose and levy tax to strengthen their economic condition. Caesar Augustus decreed that the entire world should be taxed. Statistics-2