An Improved Warner s Randomized Response Model

Similar documents
A New Mixed Randomized Response Model

On stratified randomized response sampling

Method of Estimation in the Presence of Nonresponse and Measurement Errors Simultaneously

Improved Estimation of Rare Sensitive Attribute in a Stratified Sampling Using Poisson Distribution

Improved exponential estimator for population variance using two auxiliary variables

Extension of Mangat Randomized Response Model

Estimation of the Population Mean in Presence of Non-Response

Use of Auxiliary Information for Estimating Population Mean in Systematic Sampling under Non- Response

Estimation of Population Mean Using Co-Efficient of Variation and Median of an Auxiliary Variable

Random Variables, Sampling and Estimation

A Generalized Class of Estimators for Finite Population Variance in Presence of Measurement Errors

Modified Ratio Estimators Using Known Median and Co-Efficent of Kurtosis

Improved Class of Ratio -Cum- Product Estimators of Finite Population Mean in two Phase Sampling

Chain ratio-to-regression estimators in two-phase sampling in the presence of non-response

Varanasi , India. Corresponding author

A Family of Unbiased Estimators of Population Mean Using an Auxiliary Variable

Element sampling: Part 2

Chapter 13, Part A Analysis of Variance and Experimental Design

Journal of Scientific Research Vol. 62, 2018 : Banaras Hindu University, Varanasi ISSN :

Estimation of Population Ratio in Post-Stratified Sampling Using Variable Transformation

Mathematical Modeling of Optimum 3 Step Stress Accelerated Life Testing for Generalized Pareto Distribution

Chapter 22. Comparing Two Proportions. Copyright 2010 Pearson Education, Inc.

The Sampling Distribution of the Maximum. Likelihood Estimators for the Parameters of. Beta-Binomial Distribution

Alternative Ratio Estimator of Population Mean in Simple Random Sampling

An Introduction to Randomized Algorithms

MOST PEOPLE WOULD RATHER LIVE WITH A PROBLEM THEY CAN'T SOLVE, THAN ACCEPT A SOLUTION THEY CAN'T UNDERSTAND.

A General Family of Estimators for Estimating Population Variance Using Known Value of Some Population Parameter(s)

Class 23. Daniel B. Rowe, Ph.D. Department of Mathematics, Statistics, and Computer Science. Marquette University MATH 1700

Double Stage Shrinkage Estimator of Two Parameters. Generalized Exponential Distribution

Economics Spring 2015

Math 155 (Lecture 3)

A proposed discrete distribution for the statistical modeling of

Introduction to Probability and Statistics Twelfth Edition

On the Derivation and Implementation of a Four Stage Harmonic Explicit Runge-Kutta Method *

Chapter 22. Comparing Two Proportions. Copyright 2010, 2007, 2004 Pearson Education, Inc.

AClassofRegressionEstimatorwithCumDualProductEstimatorAsIntercept

STA Learning Objectives. Population Proportions. Module 10 Comparing Two Proportions. Upon completing this module, you should be able to:

New Ratio Estimators Using Correlation Coefficient

The Random Walk For Dummies

A quick activity - Central Limit Theorem and Proportions. Lecture 21: Testing Proportions. Results from the GSS. Statistics and the General Population

SYSTEMATIC SAMPLING FOR NON-LINEAR TREND IN MILK YIELD DATA

Chapter 6 Principles of Data Reduction

Investigating the Significance of a Correlation Coefficient using Jackknife Estimates

Estimation for Complete Data

Estimation of Gumbel Parameters under Ranked Set Sampling

Enhancing ratio estimators for estimating population mean using maximum value of auxiliary variable

Properties and Hypothesis Testing

Chapter 4. Fourier Series

Chapter 8: Estimating with Confidence

Sampling Error. Chapter 6 Student Lecture Notes 6-1. Business Statistics: A Decision-Making Approach, 6e. Chapter Goals

G. R. Pasha Department of Statistics Bahauddin Zakariya University Multan, Pakistan

6.3 Testing Series With Positive Terms

Frequentist Inference

Statistical Inference (Chapter 10) Statistical inference = learn about a population based on the information provided by a sample.

Lecture Note 8 Point Estimators and Point Estimation Methods. MIT Spring 2006 Herman Bennett

4.3 Growth Rates of Solutions to Recurrences

GUIDELINES ON REPRESENTATIVE SAMPLING

A Family of Efficient Estimator in Circular Systematic Sampling

Abstract. Ranked set sampling, auxiliary variable, variance.

Hypothesis Testing. Evaluation of Performance of Learned h. Issues. Trade-off Between Bias and Variance

MBACATÓLICA. Quantitative Methods. Faculdade de Ciências Económicas e Empresariais UNIVERSIDADE CATÓLICA PORTUGUESA 9. SAMPLING DISTRIBUTIONS

If, for instance, we were required to test whether the population mean μ could be equal to a certain value μ

ON SOME DIOPHANTINE EQUATIONS RELATED TO SQUARE TRIANGULAR AND BALANCING NUMBERS

Proof of Goldbach s Conjecture. Reza Javaherdashti

Zeros of Polynomials

LESSON 20: HYPOTHESIS TESTING

A statistical method to determine sample size to estimate characteristic value of soil parameters

PH 425 Quantum Measurement and Spin Winter SPINS Lab 1

Chapter 2 Feedback Control Theory Continued

µ and π p i.e. Point Estimation x And, more generally, the population proportion is approximately equal to a sample proportion

MATH 320: Probability and Statistics 9. Estimation and Testing of Parameters. Readings: Pruim, Chapter 4

Some Exponential Ratio-Product Type Estimators using information on Auxiliary Attributes under Second Order Approximation

An Improved Proportionate Normalized Least Mean Square Algorithm with Orthogonal Correction Factors for Echo Cancellation

7-1. Chapter 4. Part I. Sampling Distributions and Confidence Intervals

Introducing Sample Proportions

FACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING. Lectures

Sect 5.3 Proportions

Confidence Intervals for the Population Proportion p

Common Large/Small Sample Tests 1/55

Seed and Sieve of Odd Composite Numbers with Applications in Factorization of Integers

Class 27. Daniel B. Rowe, Ph.D. Department of Mathematics, Statistics, and Computer Science. Marquette University MATH 1700

EXAMINATIONS OF THE ROYAL STATISTICAL SOCIETY

NANYANG TECHNOLOGICAL UNIVERSITY SYLLABUS FOR ENTRANCE EXAMINATION FOR INTERNATIONAL STUDENTS AO-LEVEL MATHEMATICS

Infinite Sequences and Series

Section 9.2. Tests About a Population Proportion 12/17/2014. Carrying Out a Significance Test H A N T. Parameters & Hypothesis

Topic 9: Sampling Distributions of Estimators

Time series models 2007

Scientific Review ISSN(e): , ISSN(p): Vol. 4, Issue. 4, pp: 26-33, 2018 URL:

ON POINTWISE BINOMIAL APPROXIMATION

This is an introductory course in Analysis of Variance and Design of Experiments.

UNIVERSITY OF TORONTO Faculty of Arts and Science APRIL/MAY 2009 EXAMINATIONS ECO220Y1Y PART 1 OF 2 SOLUTIONS

Estimation of a population proportion March 23,

Goodness-Of-Fit For The Generalized Exponential Distribution. Abstract

Average Number of Real Zeros of Random Fractional Polynomial-II

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 9

Review Questions, Chapters 8, 9. f(y) = 0, elsewhere. F (y) = f Y(1) = n ( e y/θ) n 1 1 θ e y/θ = n θ e yn

The standard deviation of the mean

Homework 5 Solutions

Topic 9: Sampling Distributions of Estimators

Maximum likelihood estimation from record-breaking data for the generalized Pareto distribution

Transcription:

Iteratioal Joural of Statistics ad Applicatios 05, 5(6: 63-67 DOI: 0.593/j.statistics.050506.0 A Improved Warer s Radomized Respose Model F. B. Adebola, O. O. Johso * Departmet of Statistics, Federal Uiversit of Techolog, Aure, Nigeria Abstract This paper presets a modificatio of Warer s [8] Radomized Respose model. Accordig to O Muircheartaigh et al [7], o-respose is ievitable i a surve; i view of this, our model further reduces the o-respose bias b further samplig for the o-respodets. I this paper we performed a empirical practice of our model ad we also performed the empirical compariso of our model with Warer [8] model. We discovered that our model is more efficiet tha the Warer [8] model. Kewords Close supervisio, Sesitive behavior, No-respodet, Sub-sample, Radomized respose techiques. Itroductio Warer [8] proposed the radomized respose techique as a surve techique to reduce potetial bias due to o-respose ad social desirabilit whe asig questios about sesitive behaviors (see Warer [8], for a comprehesive review. The method ass respodets to use a radomizatio device, such as a coi, dec of cards, spiers whose outcome is ot ow b the eumerator. The outcome of the radomized device determies which of the two questios the respodet aswers. A lot of improvemets have bee doe to Warer s radomized respose model, to metio few, Greeberg et al. [4], Gupta ad Shabbir [5], Adebola ad Adegoe [], Adepetu ad Adebola [3], Adebola et al. []. I this paper, we develop a Modificatio of Warer s Radomized Respose Techiques b itroducig the cocept of sub-sample of o-respodet. Radomized Respose Techiques helps to reduce respose ad o-respose bias while our model further reduces the o-respose bias. Item o-respose occurs whe the respodet refuses to aswer the sesitive part of the questio which is the major cocer of the iterviewer. I sectios that follow, we preset the Warer s [8] Radomized Respose Model, propose Radomized Respose Model ad thereafter its relative efficiec over the existig oe.. Warer s Radomized Respose Model * Correspodig author: teeja@gmail.com (O. O. Johso Published olie at http://joural.sapub.org/statistics Copright 05 Scietific & Academic Publishig. All Rights Reserved Warer [7] gave a geius idea b usig radomized device to ecourage truthful aswer from the respodet with respect to a sesitive behavior. The radomizig device, such as a spiig arrow, dice or cois is used to select oe of the two questios; such as, I am HI positive (class A, preseted with probabilit P I am HI egative (class B, preseted with probabilit -P The respodets have the optios Yes or No preseted to him or her. The iterviewer does ot ow which questio a respodet has aswered but ows the probabilit P ad -P with which the two statemets are preseted. Here, with a radom sample of respodets, the iterviewer records a biomial estimate θˆ of the proportio θ of Yes aswers, where is the umber of es aswers. If the questios are aswered truthfull, the relatio betwee θ ad π i the populatio is give as: πp + ( ( π ( p + ( θ π (. Where π is the proportio of people with the stigmatized or sesitive behavior usig Warer s techiques ad p is the probabilit of selectig the sesitive questio. θ (, p (. ( p Proof From equatio (., we have θ πp + ( ( π Maig π the subject of the relatio, we have θ ( p (

64 F. B. Adebola et al.: A Improved Warer s Radomized Respose Model The ubiased estimator πˆ is give b: The ariace is give b ( π of a sesitive proportio π ( ( p ˆ θ ˆ (.3 ( ( p ˆ θ ( ( p ˆ θ ( π ( p Recall that ( c 0 The, Thus, ( ( p, where c is a costat. 0 ( π ˆ θ ( p ( θ ( p θ ( π (.4 Where θˆ follows a biomial distributio, Now to fid the ubiased estimator of the variace ( p + ( ( p + p θ π θ π ( π ( ( π ( ( p π p + p p + p 4pπ + 4p π 4p π 4pπ π + π + p p p( ( p ( 4p 4p+ π ( 4p 4p+ π p p p( ( π π ( p p p The ariace is give b ( π p( ( p ( π π The ubiased variace estimator ˆ ( πˆ proportio πˆ is give b: ˆ ( ˆ π p( ( p of a sesitive ( ˆ π ˆ π (.5 The secod term i ˆ( πˆ is the variace that ˆ ( πˆ would have if all respodets aswered truthfull a direct questio about class A membership. Except b chose ππ AA ear 0.5 ad p > 0.85, the first term is greater tha the secod, ofte much greater. The method is thus quite imprecise i geeral. This might be expected sice the iterviewer does ot ow whether a es aswer implies membership i a class A or the opposite. However, Warer s method ma give a smaller mea square error (MSE tha a direct sesitive questio would, if the latter produced umerous refusals or false aswers. 3. Our Model Several radomized respose techiques has bee developed, the models developed do ot tae ito cosideratio of item o-respose (refusal to aswer the sesitive part of the questio. No-respose is a importat source of o-samplig error i surve samplig, it occurs whe some but ot all the required iformatio is collected from the sample uit. The most damagig is uit o-respose where a samplig uit refuses to aswer the sesitive part of the radomized respose techiques desiged questioaire. I view of this, we proposed a improved Warer s radomized respose model that is based o sub-sample of o-respodet so as to iduce a better estimate of the proportio of people with the stigmatized or sesitive behavior. Questioaires were set out ad the umber of useable resposes were recorded (useable resposes at the first iterview give as while the remaiig were referred to o-respose give as. I order to further reduce the o-respose bias the a surve o sub-sample of o-respodet come to place. It is assumed that the whole of the sub-sample respoded to the surve ad are useable resposes which would be achieved b close supervisio. Close supervisio i this cotext does t mea the iterviewer ows the questio aswered b the respodet but it meas the respodet is properl istructed ad moitored o a oe to oe basis. Let be the umber es respose from the respodet at the first iterview. Let m be the umber of es aswer from the sub-sample of o-respodet. Let be the sample size of the Surve. Let m be the sub-sample size. Let be the uit

Iteratioal Joural of Statistics ad Applicatios 05, 5(6: 63-67 65 which is used to tae the sub-sample. The proportio of es respose from our model is give b: m θ (3. + B simplifig, we have (see theorem for proof: ˆ θ ˆ θ + ˆ θ (3. Where θˆ be the proportio of es aswer, ˆ θ be the proportio of es aswer form the respodet at the first iterview ad ˆ θ be the proportio of es aswer from the sub-sample of o-respodet. From the Warer s radomized respose techiques, the proportio of es respose from our model is give b: θ π p+ ( ( π + π p+ ( ( π ( + π p+ ( ( π (3.3 Let B +, the we have: θ B π p+ ( ( π Solvig for π, we have, θ B( B( p The we have, θ ( + ( + p ( ( Recall that θ θ + θ θ+ θ + p + p ( ( ( ( Hece, the ubiased estimator of πˆ is give b: θ ( ( ( + ( p θ+ + p ˆ The variace of the estimator is give b: v( ˆ θ + v( ˆ θ ( ˆ π + p ( ( θ( θ θ ( θ + ( ˆ π ( + ( p B simplifig, we have (3.4 (3.5 ( ˆ π p( ( ( ( ( π p ( π ( π ( π ( + + p p + + + p + B further simplificatio, The ubiased variace estimator ˆ ( πˆ of a sesitive proportio π is give b: ( ( p ( π p ˆ π ( ˆ π + + (3.6 The, ˆ ( ˆ π ( π w + (3.7 Where ( π w is the variace for a Warer s model. Theorem : The proportio of es respose, θ is give b: ˆ θ ˆ θ + ˆ θ Proof: From the Hase ad Hurwitz [6], which itroduces the cocept of subsample of o-respodet, we have θ θ + w w θ m Where w, w, θ, θ m. m The, we have θ + (3.8 m Recall that m B substitutig for m i equatio (3.8, we have θ + The ubiased estimator of θ is give b ˆ θ ˆ θ + ˆ θ Where ˆ θ ad ˆ θ. Theorem : Give that is a iteger value ad > Show that < +

66 F. B. Adebola et al.: A Improved Warer s Radomized Respose Model Proof: Give that, That is, B re-arragig, Thus whe ( Divide through b ( + Multipl through b The, ( ( ( ( + ( ( + + From equatio (, we have > ( ( + + ( + ( ( + ( Divide through b (, we have + Subtract from both sides, we have Divide through b, we have The, From equatio ( ad (3, usig trasitivit law + 4. Comparative Stud of our Model Here we performed the comparative stud of our model; this ca be achieved mathematicall ad empiricall. Mathematicall, it follows that the proposed model is more efficiet tha the Warer s radomized respose model if we have; (3 Relative efficiec (RE ariace of proposed mod el ariace of Warer ' s mod el p( ( p( + ( p < ( π π + + p RE < π ( π B simplifig, we have RE < + Sice theorem holds, the the variace of our proposed model is less tha the variace of Warer s RRT. Empiricall, to also validate our coclusio o the proposed model we preset the tables below. Table. Table showig the relative efficiec whe 50, 0., p0.7 N π P K Warer s ariace Proposed ariace Relative Efficiec (% 50 0. 0.7 0.0056 0.009 33.33% 50 0. 0.7 3 0.0056 0.008 50% 50 0. 0.7 5 0.0056 0.0037 66.67% 50 0. 0.7 0 0.0056 0.0046 8.8% 50 0. 0.7 5 0.0056 0.0049 87.50% Table. Table showig the relative efficiec whe 500, 0., p0.7 N π P K Warer s ariace Proposed ariace Relative Efficiec (% 500 0. 0.7 0.008 0.0009 33.33% 500 0. 0.7 3 0.008 0.004 50% 500 0. 0.7 5 0.008 0.009 66.67% 500 0. 0.7 0 0.008 0.003 8.8% 500 0. 0.7 5 0.008 0.005 87.50% We ca deduce from the empirical compariso that the choice of plas a major role i the comparative stud. It ca be derived from the table that gave the miimum variace i the proposed model; coclusivel, the smaller the choice of, the more efficiet the proposed model is over the covetioal Warer s model. 5. Coclusios This paper preseted a improved Warer s radomized respose model; the proposed strateg further reduces the o-respose bias b itroducig the cocept of sub-samples of o-respodet. The proposed model is liel to iduce better estimate with a reduced variace. Moreover, the

Iteratioal Joural of Statistics ad Applicatios 05, 5(6: 63-67 67 proposed model is more efficiet tha the Warer s model. Lastl, we are able to coclude that the smaller the choice of (the uit which is used to divide the o-respodet so as to have the sub-sample size, the higher the gai i efficiec of the proposed model over the covetioal Warer s model. REFERENCES [] Adebola, F.B. ad Adegoe, N.A. (03: A Surve of Examiatio Malpractices usig the Radomized Respose Techique. Joural of the Nigeria Associatio of Mathematical Phsics, 3, 375-388. [] Adebola, F. B., Johso, O. O., & Adegoe, N. A. (04: A Modified Stratified Radomized Respose Techiques. Mathematical Theor ad Modelig, 4(3, 9-4. [3] Adepetu, A.O. ad Adebola, F.B. (04: O the Relative Efficiec of the Proposed Reparametized Radomized Respose Model. Iteratioal Joural of Mathematical Theor ad Modelig, 4, 58-67. [4] Greeberg, B.G., Abul-Ela, A.A., Simmos, W.R. ad Horvitz, D.G. (969: The Urelated Questio Radomized Respose: Theoretical Framewor. Joural of the America Statistical Associatio 64, 50-539. [5] Gupta, S.N. ad Shabbir, J. (006: A Alterative to Warer s Radomized Respose Model. Joural of Moder Applied Statistical Methods, 5, 38-33. [6] Hase, M.H. ad Hurwitz, W.N. (946: The Problem of No-Respose i Sample Surves, Joural of the America Statistical Associatio 4, 57-59. [7] O Muircheartaigh, C. ad Campaelli, P. (999: A multilevel exploratio of the role of iterviewers i surve o-respose. Joural of the Roal Statistical Societ, Series A 6, 437-46. [8] Warer, S.L. (965: Radomized respose: A Surve Techique for Elimiatig Evasive Aswer Bias, Joural of the America Statistical Associatio 60, 63-69.