Global Adult Tobacco Survey (GATS) Sample Design Manual. Version 2.0 November 2010

Transcription

1

2

3 Global Adult Tobacco Survey (GATS) Sample Deign Manual Verion 2.0 November 2010

4 Global Adult Tobacco Survey (GATS) Comprehenive Standard Protocol GATS Quetionnaire Core Quetionnaire with Optional Quetion Quetion by Quetion Specification GATS Sample Deign Sample Deign Manual Sample Weight Manual GATS Fieldwork Implementation Field Interviewer Manual Field Supervior Manual Mapping and Liting Manual GATS Data Management Programmer Guide to General Survey Sytem Core Quetionnaire Programming Specification Data Management Implementation Plan Data Management Training Guide GATS Quality Aurance: Guideline and Documentation GATS Analyi and Reporting Package Fact Sheet Template Country Report: Tabulation Plan and Guideline Indicator Definition GATS Data Releae and Diemination Data Releae Policy Data Diemination: Guidance for the Initial Releae of the Data Tobacco Quetion for Survey: A Subet of Key Quetion from the Global Adult Tobacco Survey (GATS) Suggeted Citation Global Adult Tobacco Survey Collaborative Group. Global Adult Tobacco Survey (GATS): Sample Deign Manual, Verion 2.0. Atlanta, GA: Center for Dieae Control and Prevention, ii

5 Acknowledgement GATS Collaborating Organization Center for Dieae Control and Prevention CDC Foundation John Hopkin Bloomberg School of Public Health RTI International Univerity of North Carolina Gilling School of Public Health World Health Organization Financial Support Financial upport i provided by the Bloomberg Initiative to Reduce Tobacco Ue, a program of Bloomberg Philanthropie, through the CDC Foundation. Diclaimer: The view expreed in thi manual are not necearily thoe of the GATS collaborating organization. iii

6 iv

7 Content Chapter Page 1. Introduction Overview of the Global Adult Tobacco Survey Ue of thi Manual Deciding on a GATS Sample Deign Survey Deign Objective Target Population and Sample Frame Definition of the GATS Target Population and Study Eligibility Area Sampling Frame Baic Survey Deign Specification Sample Deign Feature Sample Size and Expected Preciion Data Collection Procedure Overview of the Sampling Approach Forming Primary Sampling Unit (PSU) Firt Stage of Sampling: Selecting PSU Countrie Selecting Geographic Area (i.e., Segment) in One Stage of Selection Countrie Selecting Geographic Area in Several Stage of Selection Selecting Houehold Creating a Houehold Liting for a Sampled Segment by Viiting the Segment Prior to Selecting the Houehold Sample Creating a Houehold Liting for a Sampled Segment Uing Adminitrative Record Selecting the Sample of Houehold Optionally Aigning Houehold to the Male and Female Houehold Group Preparing an Electronic Sample File for Handheld Computer Adminitration Addreing Mied Houehold and Other Apparent Error in the Houehold Liting Selecting an Eligible Reident Within Each Screened Houehold Determining Sample Size at Each Stage of Selection and Reporting Final Repone Rate Sample Size at Each Stage of Selection Example Sample Size Computation Repone Rate Computing Final Sample Weight for Analyi v

8 12. Bibliography Sampling: Simple and Brief Sampling: Introductory Sampling: More Advanced Treatment Practical Manual Other Paper GATS Manual Appendix A. Rationale for the GATS Recommended Sample Size... A-1 Appendix B. Sample Deign Option for New GATS Countrie... B-1 Appendix C. Sample Deign Option for Repeat GATS Countrie... C-1 vi

9 Exhibit Number Page Exhibit 3-1. GATS Reidency Example Exhibit 5-1. Overview of GATS Sample Deign Exhibit 8-1. Summary of Proce to Addre Mied Houehold on Houehold Liting Exhibit Houehold Quetionnaire Final Dipoition Code Exhibit Individual Quetionnaire Final Dipoition Code vii

10 viii

11 1. Introduction Tobacco ue i a major preventable caue of premature death and dieae worldwide. Approximately 5.4 million people die each year due to tobacco-related illnee a figure expected to increae to more than eight million a year by If current trend continue, tobacco ue may kill a billion people by the end of thi century. It i etimated that more than three quarter of thee death will be in low- and middle-income countrie 1. An efficient and ytematic urveillance mechanim i eential to monitor and manage the epidemic. The Global Adult Tobacco Survey (GATS), a component of Global Tobacco Surveillance Sytem (GTSS), i a global tandard for ytematically monitoring adult tobacco ue and tracking key tobacco control indicator. GATS i a nationally repreentative houehold urvey of adult, 15 year of age or older, uing a tandard core quetionnaire, ample deign, and data collection and management procedure that have been reviewed and approved by international expert. GATS i intended to enhance the capacity of countrie to deign, implement and evaluate tobacco control intervention. In order to maximize the efficiency of the data collected from GATS, a erie of manual ha been created. Thee manual are GATS manual provide ytematic deigned to provide countrie with tandard requirement a well guidance on the deign and a everal recommendation on the deign and implementation of implementation of the urvey. the urvey in every tep of the GATS proce. They are alo deigned to offer guidance on how a particular country might adjut feature of the GATS protocol in order to maximize the utility of the data within the country. In order to maintain conitency and comparability acro countrie, following the tandard protocol i trongly encouraged. 1.1 Overview of the Global Adult Tobacco Survey GATS i deigned to produce national and ub-national etimate among adult acro countrie. The target population include all non-intitutionalized men and women, 15 year of age or older, who conider the country to be their uual place of reidence. All member of the target population will be ampled from the houehold (HH) that i their uual place of reidence. GATS ue a geographically clutered, multitage ampling methodology to identify the pecific houehold that Field Interviewer will contact. Firt, a country i divided into Primary Sampling Unit, egment within thee Primary Sampling Unit, and houehold within the egment. The GATS interview i compoed of Then, a random ample of houehold i elected to participate in two part: Houehold Quetionnaire GATS. and Individual Quetionnaire. Thee The GATS interview conit of two part: the Houehold Quetionnaire (houehold creening) and the Individual quetionnaire are adminitered uing an electronic data collection device. 1 Mather, C.D., and Loncar D. Projection of Global Mortality and Burden of Dieae from 2002 to PLoS Medicine, 2006, 3(11):e442. Global Adult Tobacco Survey (GATS) 1-1 Sample Deign Manual Chapter 1: Introduction

12 Quetionnaire (individual interview). Both interview are conducted uing an electronic data collection device. At each addre in the ample, Field Interviewer will adminiter the Houehold Quetionnaire to one adult who reide in the houehold. The purpoe of the Houehold Quetionnaire are to determine if the elected houehold meet GATS eligibility requirement and to make a lit, or roter, of all eligible member of the houehold. Once the roter of eligible reident of the houehold i completed, one individual will be randomly elected to complete the Individual Quetionnaire. The Individual Quetionnaire ak quetion about background characteritic; tobacco moking; mokele tobacco; ceation; econdhand moke; economic; media; and knowledge, attitude, and perception about tobacco. 1.2 Ue of thi Manual Thi manual i deigned to offer both requirement and recommendation, a well a uggeted guideline for countrie to follow a they develop an appropriate ample deign for their implementation of GATS. Thi chapter provide background information on GATS. Subequent chapter are ummarized below: Chapter 2 ummarize the urvey objective. Chapter 3 preent a definition of the target population for the urvey and provide dicuion of the area ample frame that will be ued. Clearly defining the target population and ample frame i imperative. One can better undertand the extent of ample coverage of the target population when a clear definition of the target population i etablihed. Thi definition alo will be ued to determine urvey eligibility for ampling and data collection. Thi chapter alo provide a definition of a houehold 2. Chapter 4 preent the baic feature of the ample deign that are required to achieve betweencountry comparability a well a etablihed level of tatitical quality. Other deign requirement and recommendation are alo preented in the following chapter. Chapter 5 preent a brief overview of the recommended approach for GATS ample. Chapter 6-9 preent a ummary of the ample election proce at each tage of the ample deign. Chapter 10 offer uggetion on determining an adequate ample ize to elect at each tage of the deign. Chapter 11 provide a brief overview of the proce of producing a ample weight for each repondent. Thee weight are typically adjuted for nonrepone and calibrated to target population count before they are ued for data analyi. 2 It hould be noted that while we ue the term houehold and houing unit (or dwelling unit ) omewhat interchangeably in thi manual, they have omewhat different meaning. A houehold i peron or group of people who live together in a houing unit. Thu, houing unit refer the phyical tructure (e.g., ingle-family home, apartment, mobile home, group of room, etc.) in which the peron or peron compriing the houehold reide. Global Adult Tobacco Survey (GATS) 1-2 Sample Deign Manual Chapter 1: Introduction

13 Chapter 12 provide a lit of reference that might be ueful to tatitician a they develop the pecific GATS ample deign for their country. Appendix A preent the rationale for the recommended overall ample ize for GATS urvey. Appendix B offer a general et of N Option for the ample deign in any of the new GATS countrie after the iuance of thi manual econd edition in Appendix C offer a general et of R Option for countrie that have completed a urvey round of GATS and are about to conduct a repeat round. 1.3 Deciding on a GATS Sample Deign Thi manual provide a general tatitical template for the probability ample that hould be ued for any GATS. The purpoe of thi ection i to lay out everal baic option for each of two type of deign etting that are relevant at thi point in the development of the GATS ytem of urvey in which everal countrie have already done a GATS and other are looking to complete a GATS for the firt time. The two type of deign etting are: (i) new countrie hoping to conduct their firt round of GATS, and (ii) countrie wihing to repeat GATS by conducting another urvey round. Sample Deign Option for Countrie Conducting Firt Round of GATS The deign option offered to new countrie are delineated according to: (i) whether or not the national ample ued for GATS i tatitically derived from an exiting national ample (i.e., whether or not the GATS ample deign i integrated with an exiting ample deign), and (ii) whether or not ample ize need to be ufficient to (beide meauring gender difference) alo make urban-rural comparion. A GATS deign that i integrated with a highly regarded exiting ample and data collection infratructure may imply lower urvey cot than a tand-alone ample deign (i.e., one that i developed excluively for GATS). Moreover, a GATS ample deigned only to adequately meaure difference between two gender categorie can be maller than a ample for which one wihe to have adequately precie etimate for four population ubgroup formed by the cro-claification of gender by urbanicity. The following three deign option conidered in thi etting are: OPTION N1 with a tandard tand-alone deign and overall ample ize of 8,000, which i ufficient to produce etimate for the four gender-by-urbanicity population ubgroup. OPTION N2 with a maller tand-alone deign and overall ample ize of 4,000, which i ufficient to produce etimate for the two gender group and the two urbanicity ubgroup eparately, but not for the four gender-by-urbanicity population ubgroup. OPTION N3 with a ample deign that i integrated with a highly regarded exiting national ample deign and whoe repondent ample ize may be at the tandard level (8,000) or maller. See Appendix B for more detail on thee option. Sample Deign Option for Countrie Repeating GATS For countrie repeating GATS, the focu i on how the ample for the upcoming (repeat) round relate to the ample from the initial (baeline) round. The option here are delineated according to: (i) whether or Global Adult Tobacco Survey (GATS) 1-3 Sample Deign Manual Chapter 1: Introduction

14 not the two ample have any planned overlap, and (ii) whether or not the tructure (i.e., defined by the ampling unit in each tage and tratum definition) of the deign for the repeat and baeline ample are the ame. The following three deign option conidered in thi etting are: OPTION R1 where there i partial (planned) overlap between the ample in the baeline and repeat round. OPTION R2 where the tructure of the baeline and repeat round ample are identical but the two ample are choen independently. OPTION R3 where the tructure of the baeline and repeat round ample are not identical and the two ample are choen independently. See Appendix C for more detail on thee option. Global Adult Tobacco Survey (GATS) 1-4 Sample Deign Manual Chapter 1: Introduction

15 2. Survey Deign Objective Survey deign requirement and recommendation for GATS have been developed o that high quality etimate can be generated for each country a a whole a well a for two analyi group defined by urbanicity and gender. A will be noted in Chapter 4, the targeted repondent ample ize i typically defined o that approximately half of the repondent ample will reide in urban area and half in rural area. Thi allocation of the ample will be accomplihed by explicitly tratifying the ample by urbanicity (i.e., urban v. rural) at the earliet poible tage() of the deign. A an option, the ample deign ha alo been defined o that the mixture of male and female in the repondent ample can be controlled by randomly aigning elected houehold to male or female gender group through a proce called gender randomization. For example, if a houehold i randomly aigned to the male group, then only eligible male reident will be lited and elected from the houehold for interview. Similarly, if the houehold i randomly aigned to the female group, then only eligible female reident will be lited and elected from the houehold. Gender randomization i done to accomplih either or both of the following: (i) to predetermine the gender of elected houehold member o that matching the gender of the field interviewer and repondent can be eaily accomplihed in culture where a gender mimatch hould be avoided, or (ii) to diproportionately ample male or female depending on ample ize need by gender. Gender randomization will be optional to GATS countrie. Therefore, a etting will be intalled in the program code to enable countrie to etablih whether or not gender-pecific houehold roter will be created in participating ampled houehold. The default option for thi programming parameter will be to identify eligible houehold reident a if gender randomization had not been done. Deigning the urvey to generate precie cro-ectional etimate by gender and urbanicity i preferred for GATS, primarily to allow Recommendation: comparion of etimate by thee domain between different The deign for the tandard verion countrie participating in thi effort. Some countrie may have other of GATS eek to obtain precie high-priority demographic domain for reporting urvey etimate. etimate jointly by urbanicity For example, a country may be intereted in generating precie and gender. etimate for ome geographic region where region may be individual or group of tate, province, or any ort of geographic entitie defined by landmark, political belief or other cultural factor. Thee type of additional data requirement impoed by a country are acceptable, provided etimate of acceptable tatitical preciion can be produced for them and they do not compromie the tatitical quality of etimate for major domain of interet in intra- and inter-country analye. It i worth noting that the ucce of generating etimate by geographic region, in particular, generally implie that overall ample ize mut be ubtantially larger than if no regional etimate are required. In later ection of thi manual it will be noted that the GATS ample hould be deigned to obtain a minimum number of 8,000 repondent per country with roughly half the ample allocated to urban area and half to rural area. If any country i intereted in obtaining regional etimate and would like to compare etimate between urban and rural area within each region, it i recommended that the urvey be deigned to obtain 8,000 repondent in each region (half allocated to the urban area and half to the Global Adult Tobacco Survey (GATS) 2-1 Sample Deign Manual Chapter 2: Survey Deign Objective

16 rural area). If a country i intereted in obtaining regional etimate and i not intereted in comparing etimate between urban and rural area of the region, then it i recommended that the urvey be deigned to obtain 4,000 repondent in each region. Thee minimum ample ize will generate etimate that meet the expected preciion requirement of the GATS. Global Adult Tobacco Survey (GATS) 2-2 Sample Deign Manual Chapter 2: Survey Deign Objective

17 3. Target Population and Sample Frame 3.1 Definition of the GATS Target Population and Study Eligibility In general, the target population of the GATS hould include all eligible individual reiding in all geographic area within a GATS Only in rare intance hould countrie country. In rare intance it may be neceary to exclude ome exclude geographic portion of the geographic area. Areal excluion will be conidered, provided a country from the GATS target ignificant proportion of the country population doe not reide in population. thee area. For example, extremely remote area or area that cannot be viited due to war, political unret, etc. may be conidered for excluion from the GATS target population. To propoe an excluion, a country hould (i) explicitly define which geographic area of the country are included in and excluded from the target population, and (ii) provide an etimate of the percent of the population that reide in thoe area being excluded. Thi percentage will provide an indication of the potential bia effect of any areal excluion on etimate that are otherwie intended to be national in cope. The target population for GATS urvey hould include the civilian, non-intitutionalized population of men and women, 15 year of age or older, and living in the included area of the country, who meet GATS reidency requirement both at the country and houehold level. Thi definition i important when completing the Houehold Quetionnaire, which include a ection where all eligible member of the houehold are lited and one i randomly choen to complete the Individual Quetionnaire. Eligible reident of jut the deignated gender group for the houehold hould be lited if gender randomization of houehold i done. GATS Target Population: Include all civilian, nonintitutionalized men and women 15 year of age or older, who live in the included area of the country, and who meet GATS reidency requirement for the country a well a for the houehold where they are ampled. Individual, 15 year of age or older, who are explicitly excluded from the urvey, are thoe who, at the time that the Houehold Quetionnaire i completed, are: non-citizen viiting the country for a few week (e.g. tourit, in the country to ee friend/ relative, etc.); citizen in the military who indicate that their uual place of reidence i either on or off of a military bae, OR citizen who are intitutionalized including people reiding in hopital, prion, nuring home, and other uch intitution, who will not be ampled in GATS. Country Reidence Requirement for GATS: Individual are viewed a reident of a GATS country if they are: (i) citizen of and reiding in the country, or (ii) non-citizen living in the country, but who conider the country to be their uual country of reidence (i.e., they have lived in the GATS country for at leat half of the time during the 12 month prior to completing the Houehold Quetionnaire). Global Adult Tobacco Survey (GATS) 3-1 Sample Deign Manual Chapter 3: Target Population and Sample Frame

18 Houehold Reidence Requirement for GATS: All civilian, non-intitutionalized men and/or women, 15 year of age or older, who are enrolled in a chool and live in a dormitory while the chool i in eion, but who conider the elected houehold to be their place of reidence otherwie, automatically meet the houehold reidence requirement for GATS 1. Some non-citizen of a GATS country may meet the country reidence requirement. Non-citizen qualify a uual reident of a GATS country if they have reided in that country for at leat half of the time during the pat 12 month. All other civilian, non-intitutionalized men and women, 15 All tudent living away in dormitorie meet the year of age or older, in a ampled houehold meet the houehold reidence requirement for GATS. houehold reidence requirement if the ampled houehold i conidered to be their uual place of reidence at the time that the Houehold Quetionnaire i completed. A ampled houehold i the uual place of reidence for an otherwie-eligible peron living there, if that peron either ha no other reidence, or ha multiple reidence but ha lived at the ampled houehold for at leat half of the time in the pat 12 month. In addition, a peron who ha recently moved to the ampled houehold to make it hi/her ole reidence i conidered to be a member of that houehold if he/he doe not plan to return to hi/her previou houehold. Converely, a peron who recently moved out of the ampled houehold, and ha no known intention of returning, i no longer conidered a member of the ampled houehold. Except for tudent living in dormitorie, the uual place of reidence for houehold member with more than one A uual member of a ampled houehold i any reidence i the place where they pent at leat half of otherwie-eligible reident who: (i) ha no other the time in the pat 12 month. Procedurally, it i reidence, or (ii) ha multiple reidence but ha important to note that if the reident randomly elected been living in the elected houehold for at leat to complete an Individual Quetionnaire are temporarily half of the time during the pat 12 month. away from the houehold at the time of election, every poible effort hould be made to complete the interview with them at a later time. They hould not be arbitrarily (or even randomly) replaced by another eligible member of the houehold. Exhibit 3-1 preent ome pecific example regarding application of country and houehold reidency requirement in completing houehold roter of non-intitutionalized men and/or women, 15 year of age or older. It include an indication a to whether or not they would be eligible for election to complete an Individual Quetionnaire. 1 Since chool dormitorie will not be included on the lit ued to elect houehold, but tudent temporarily living away from home are included in the GATS target population, we mut conider peron living in thee dormitorie to be reident of the houehold of which they are a part where they are not in chool. Global Adult Tobacco Survey (GATS) 3-2 Sample Deign Manual Chapter 3: Target Population and Sample Frame

19 Exhibit 3-1. GATS Reidency Example Example of peron for whom country and houehold reidency mut be determined A citizen i a tudent currently living in a dormitory at chool, but he return to the ampled houehold when chool i not in eion. A citizen i a tudent and he live in a chool dormitory during the chool year, but he i at home on vacation. A temporary farm worker i a citizen who conider the ampled houehold to be home, but he ha been working elewhere in the country for nine of the pat 12 month. A foreign citizen whoe only reidence ha been the ampled houehold for the pat 10 month while he complete a temporary work aignment for hi foreign employer. A citizen ha jut moved into her newly purchaed ole reidence, which i the ampled houehold. At the time that the houehold roter i completed, a citizen i living in her vacation home, which wa elected a the ampled houehold and where he ha tayed two day per week in the pat year. A citizen own two home and the one elected for GATS i her primary reidence where he ha lived all but one month during the pat year; however, he i living at the other (vacation) home at the time that the roter i completed. Should the interviewer conider thi peron eligible to be elected for the Individual Quetionnaire and include thi peron on the houehold roter? 2 Ye. She meet the country eligibility criterion; and ince dormitorie are not ampled, the only way for thi peron to have a chance to be elected i to conider her to be a reident of the ampled houehold. Ye, and for the ame reaon a in the previou example. No. Thi peron meet the country criterion ince he i a citizen, but he ha multiple reidence and ha not lived at the ampled reidence more than half of the time during the pat year. Ye. Thi peron doe meet the country reidence criterion, ince he ha lived in the country for at leat half of the time in the pat year; he alo meet the houehold reidence criterion ince thi i hi only reidence. Ye. She meet the country reidency criterion, and intend to tay at what would be conidered her only reidence, thu enabling her to meet the houehold reidence criterion. No. She i a citizen and thu meet the country reidence criterion, but he ha multiple reidence and ha not lived at the ampled reidence more than half of the time during the pat year. Ye. She i a citizen and thu meet the country reidence criterion, and he ha multiple reidence and ha lived at the ampled houehold more than half of the time during the pat year. In ummary, at the initial viit made by the interviewer to each elected houehold, the interviewer will create a roter of all eligible reident who conider the elected houehold to be their uual place of reidence at the time the roter i completed. Moreover, all eligible houehold reident hould be included on the roter, and all reaonable effort hould be made to interview them if they are elected to complete the Individual Quetionnaire. The GATS Field Interviewer Manual will provide country-pecific information on who i included and excluded from the target population, a well a a dicuion of when to follow-up with elected individual who are currently not reiding in their uual place of reidence. 2 Aume that the reference peron in each example meet all other eligibility criteria for GATS. Global Adult Tobacco Survey (GATS) 3-3 Sample Deign Manual Chapter 3: Target Population and Sample Frame

20 3.2 Area Sampling Frame In general, the ampling frame for a urvey ample i the lit of all ampling unit in the urvey population from which the ample i to be drawn. There are two type of frame ued in mot face-to-face urvey of human ubject: lit frame and cluter (e.g., area) frame. A lit frame i a imple lit of population member (e.g., a lit of tudent in a city chool; a lit of file to be randomly elected from a room full of file cabinet). A cluter frame i a lit of entitie that provide indirect reference to individual by grouping them omehow (e.g., a lit of city chool, each being a grouping of tudent; or a lit of the room file drawer, each containing many file). Thi method i called indirect becaue a ample of cluter mut be elected firt, and then the ample of individual i identified from the individual linked to the elected cluter. For the GATS, a with mot population tudie, a comprehenive lit of all urvey-eligible individual in a country imply doe not exit, thu making ampling from a lit frame virtually impoible. Conequently an indirect cluter ample frame mut be ued to elect the ample. In ummary, geographic area within each country will be randomly elected, a complete lit of all houehold within each elected area will be contructed, a ample of houehold will be randomly elected from each lit, and finally one individual will be randomly elected from each elected houehold for interview. Thi type of deign aume that each elected urvey-eligible individual can be linked to one, and only one, houehold in the country. Definition of a Houehold The previou ection provided a formal definition of a urvey individual for the GATS. Equally important i to provide a formal definition of a houehold. While thi definition can be changed omewhat by a country, in general we ugget that countrie ue the definition of a houehold a defined by the United Nation in their report Principle and Recommendation for Population and Houing Cenue. The report define a houehold 3 a: Either a one-peron houehold, defined a an arrangement in which one peron make proviion for hi or her own food or other eential for living without combining with any other peron to form part of multi-peron houehold or a multi-peron houehold, defined a a group of two or more peron living together who make common proviion for food or other eential for living. The peron in the group may pool their income and have related or unrelated peron or a combination of peron both related and unrelated. Thi arrangement exemplifie the houekeeping concept. In an alternative definition ued in many countrie exemplifying the o-called houehold-dwelling concept, a houehold conit of all peron living together in a houing unit. The complete, final definition of a houehold that a country plan to ue hould be included in the country GATS ample deign propoal, and later, in it report of GATS finding. 3 Thi definition of a houehold wa taken from Global Adult Tobacco Survey (GATS) 3-4 Sample Deign Manual Chapter 3: Target Population and Sample Frame

21 4. Baic Survey Deign Specification The ampling unit in each tage of election refer to the entitie that are elected for the urvey. In thi urvey, the ultimate ampling unit are the houehold and one individual reiding within the elected houehold. In general, the ample for GATS ha been deigned to be elected uing a multi-tage, geographically clutered deign. A multi-tage deign i defined a a ample deign that entail progreively electing ubample from a previouly elected ample until the ultimate ampling unit are elected. In the GATS, exiting geo-political area will be elected at the early tage of the deign; houehold, and ultimately individual within houehold, will be elected at the latter tage of the deign. Additional detail of the deign are preented in the next ection. A noted in Chapter 1, there are certain requirement and recommendation that hould be followed in order to maximize the comparability of the reult between countrie that are conducting GATS. However, each country ha the option of introducing deign enhancement that would allow them to increae the uability of the reult from thi urvey (e.g., electing the ample to enure precie etimate by region). In thi chapter, we preent ome of the baic urvey deign requirement. Any deign enhancement that a country wihe to introduce will generally be acceptable provided it doe not interfere with thee baic requirement. Additional pecific requirement on the ample deign are preented in the remaining ection of thi manual. In ummary, ome of the required baic urvey deign feature of the GATS are preented in Section 4.1 to Sample Deign Feature Requirement related to the ample deign include the following: 1. Random election mut be ued in each ampling tage o that every member of the target population ha a non-zero chance of being elected into the ample. REQUIREMENT: At each tage of election, every unit mut be elected randomly and probabilitie of election at each tage mut be recorded. 2. The probability of election for every unit (houehold and individual) elected at each tage of the deign mut be known and retained on the final analytic file for the urvey. For example, if the ample i elected in four tage (primary ampling unit, or PSU, election; egment 1 election; houehold election; and individual election) then the final analytic file mut contain the correponding four probabilitie of election. At a minimum, each country hould create and maintain two file: a. Houehold-Level File Thi file will have a record for every houehold elected for GATS. Each record hould include a unique houehold identification number; the geographic probabilitie of election (e.g. the PSU and egment probabilitie of election); and the probability of electing the houehold within each elected egment, all trata variable, and cluter identifier. An indicator of the reult of 1 A egment i defined a a maller geographic area within a particular primary ampling unit. One or more intermediate ampling tage may be required to elect egment within PSU. Global Adult Tobacco Survey (GATS) 4-1 Sample Deign Manual Chapter 4: Baic Survey Deign Specification

22 gender allocation (to male or female group), along with the correponding probability of allocation to the allocated gender group, hould alo be included in thi file, if gender allocation i ued by the country. Finally, countrie may wih to conider creating a flag variable (derived from Houehold Quetionnaire final dipoition code) indicating whether the elected houehold wa eligible or not for GATS, and a econd flag variable indicating whether or not the houehold ha completed the houehold roter. b. Peron-Level File Thi file will have a record for every individual elected for interview from uccefully creened houehold. Each record hould include all the variable in the houehold file, plu a unique individual identifier and the probability of electing the individual. Additionally, countrie may wih to conider creating a flag variable (derived from Individual Quetionnaire dipoition code) indicating whether the elected individual wa eligible or not for the GATS and a econd flag variable indicating whether or not, baed on the final dipoition code for the individual, the elected individual ha completed the Individual Quetionnaire. Additional information on thee variable i provided in ubequent ection of thi manual. 3. Countrie hould create unique identifier for every ampling unit elected at each tage of the deign, and thee identifier REQUIREMENT: mut be retained in the final analytic file. For example, if a Create unique identifier for every ample of individual i elected in four tage, then the ample unit at each election tage. analytic file hould contain unique identifier at the PSU, egment, houehold and individual level. Thee identifier are needed to verify the probability of election at each tage of the deign and may alo be needed to etimate variance of the etimate produced from the urvey. Except for firt tage ampling unit and tratification identifier, countrie hould not make thee identifier available in a public ue file in order to preerve the confidentiality of each repondent in the urvey. 4. A noted earlier, the ample hould be elected uing a multi-tage, geographically clutered ample deign. Detail of thi type of ample deign will be provided in the next ection of thi manual. It will be noted in thee ection that for the firt tage of election, countrie hould partition their country into geographic, compact, non-overlapping area. Thee area will be referred to a the primary ampling unit, or PSU, for the GATS. Thee area can be formed uing ome natural, political, economic or other uch geographic boundarie uch a tate, countie, province, village, or cenu ector. Note that one of the main reaon for clutering in a ample deign i to minimize travel requirement of the data collector by concentrating the work in geographic pocket, or cluter. 5. In general, there hould be at leat houehold within each PSU (ome countrie may have coniderably more), and REQUIREMENT: at leat 100 PSU hould be elected at the firt tage of the Select at leat 100 PSU, regardle deign. Thi minimum PSU ample ize hould be ufficient to of the overall repondent ample ize. provide adequate geographic repreentation of a country while till offering the deired reduction in data collection cot that are ought in a clutered deign. It i important to keep the number of ample PSU a large a Global Adult Tobacco Survey (GATS) 4-2 Sample Deign Manual Chapter 4: Baic Survey Deign Specification

23 poible ince the maller the PSU ample ize, the larger the average repondent ample ize per PSU, and the relatively le precie urvey etimate will be compared to an unclutered repondent ample of the ame ize. 6. The number of PSU that a country i partitioned into at the firt tage of the deign hould be large enough (>1,000) o that the ampling rate for the PSU election will not be greater than about 10%. Large geographic region can be ued to form trata, but they hould not be ued a PSU. Additional information on forming PSU i provided in Chapter 6 of thi manual. 4.2 Sample Size and Expected Preciion Requirement and recommendation related to other meaure of ample ize are baed on the following indicator of tatitical quality that were etablihed for GATS finding: 1. GATS hould be deigned to produce etimate that meet the following preciion requirement: Etimate computed at the national level, by urbanicity, by gender and by the cro of gender and urbanicity hould have a 95% confidence interval with a margin of error of 3 percentage point or le for tobacco ue rate of 40%. Sample ize for the initial round of the GATS, a well a poibly ubequent round, 2 hould be ufficiently large to accommodate the following requirement for tet to detect difference between urvey round baed on independently choen ample at each round: Tobacco ue prevalence rate reduced from 40% to 34% with 80% power and to 33% with 90% power, Type I error of 0.05, and a two-ided alternative. Secondhand expoure rate reduced from 80% to 70% with >90% power, Type I error of 0.05, and a two-ided alternative. Quit attempt rate increaed from 10% to 20% with >90% power, Type I error of 0.05, and a two-ided alternative. 2. The deign effect aociated with any particular etimate from a urvey i defined a the inflation of variance due to complex urvey deign feature uch a unequal weighting and clutering. Mathematically, it i the ratio of the variance of an etimate baed on the complex urvey deign relative to the correponding variance of the ame ample ize uing imple random ampling. While it i theoretically poible to achieve a deign effect le than one, in practice the complex deign feature of a urvey nearly alway ha a detrimental effect on preciion of the etimate. Therefore, for mot tudie, the deign effect will be greater than one. Recommendation: A repondent ample ize of 2,000 i recommended for each key reporting domain to meet GATS tandard of tatitical quality et for the domain etimate. Thi correpond to a recommended overall repondent ample ize of at leat 8,000 for national etimate that are to be jointly reported by gender and urbanicity. Interview at leat 8,000 repondent for country-level etimate by urbanicity and gender. 2 Note that funding for GATS i on a round-by-round bai. Global Adult Tobacco Survey (GATS) 4-3 Sample Deign Manual Chapter 4: Baic Survey Deign Specification

24 Auming a deign effect of 2.00 for etimate computed at the national level, by urbanicity, by gender, and by the cro of gender and urbanicity, the minimum ample ize needed to attain the GATS tandard of tatitical quality jut decribed i 2,000 repondent. When applied to each of the four group defined by the cro of urbanicity and gender, thi reult in a minimum recommended repondent ample of 8,000, which i et a the recommended overall ample ize when the cro-claification of gender-by-urbanicity at the national level define the key reporting domain for GATS finding. Additional detail on how thi repondent ample ize wa computed i provided in Appendix A. 3. If any country i intereted in obtaining regional etimate and Recommendation: would like to produce etimate jointly by gender and urbanicity A repondent ample ize of at leat in each region, then the urvey hould be deigned to obtain 8,000 i recommended for each region 8,000 repondent in each region (half allocated to the urban when etimate are to be reported area and half to the rural area). If a country i intereted in jointly by gender and urbanicity in each obtaining regional etimate, but only wihe to report etimate eparately by gender and urbanicity, then the urvey region. When etimate are to be hould be deigned to obtain 4,000 repondent in each region. reported eparately by gender and urbanicity in each region, 4, If any country introduce enhancement that would reult in the expected deign effect to generally exceed 2.00, or ha repondent per region i ufficient. empirical data that indicate that mot deign effect are likely to be greater than 2.00, then the ample ize within the group hould be appropriately adjuted upward. For example, if a country aume a deign effect of 2.5, then their deign mut be adjuted to yield: ,000 = 10,000 repondent The deign of the urvey hould correctly reflect anticipated level of nonrepone and ineligibility in determining how many REQUIREMENT: houehold mut be elected in order to yield the GATS ample deign hould addre recommended number of repondent indicated above. potential nonrepone and ineligibility Nonrepone and ineligibility may be oberved at both the at each tage. houehold level and the elected individual level. For example, a houehold adult reident aked to complete a roter of eligible reident, or a peron elected for interview, may refue to participate (nonrepone). Similarly, a elected houehold may prove to be vacant or a elected peron may indicate they are le than 15 year old and therefore ineligible. A more comprehenive definition of nonrepone and ineligibility can be found in the GATS Field Interviewer Manual. Global Adult Tobacco Survey (GATS) 4-4 Sample Deign Manual Chapter 4: Baic Survey Deign Specification

25 If, for example, a country urvey i deigned to achieve 8,000 repondent and it expect to oberve the following: Rate Comment Aumption Houehold Eligibility Rate Ineligible houehold include thoe that are vacant. 90% Houehold Screening Rate Account for houehold with no adult, 15 year of age or older. If gender randomization i ued, thi alo account for houehold elected for the male houehold group and no male reide in the houehold. Similarly thi account for houehold elected for the female houehold group and no female reide in the houehold. 95% Houehold Repone Rate Account for eligible creened houehold where a Houehold Quetionnaire (the houehold roter) i not completed. 98% Individual Eligibility Rate In ome cae, an individual might be interviewed for GATS and later determined to be ineligible; e.g., they are in fact, younger than 15 year old. Thi rate account for thi. 98% Individual Repone Rate Account for thoe eligible houehold reident who are elected but do not complete the Individual Quetionnaire. 85% Individual Repone Rate for Female An individual repone rate when the houehold i randomly allocated to be a female houehold. 90% Individual Repone Rate for Male An individual repone rate when the houehold i randomly allocated to be a male houehold; pat tudie ugget male will repond at lightly lower rate than female. 80% Then, in thi example, the urvey mut be deigned to achieve a elected ample of 4,000 = 4, ,000 = 5, ,535+ 5,102 = 11, elected female, elected male, and elected houehold. Global Adult Tobacco Survey (GATS) 4-5 Sample Deign Manual Chapter 4: Baic Survey Deign Specification

26 Additional guideline for determining an appropriate ample ize at each tage of the GATS ample deign are provided in Chapter Data Collection Procedure Requirement related to the data collection procedure include the following: 1. Each country mut collect the data via a face-to-face data collection method. In addition, each country hould ue the core part of the GATS quetionnaire that i preented in the GATS Core Quetionnaire and Optional Quetion manual. REQUIREMENT: Obtain combined repone rate of at leat 80%. 2. Each country mut develop data collection procedure to achieve a combined repone rate of 80% or greater. At each tage of the deign, the repone rate i defined a the total number of repondent divided by the total number of eligible individual. The combined repone rate i defined a the product of the tage-pecific repone rate. In GATS then, the combined repone rate i defined a the product of the houehold and individual-level repone rate. Specific information on defining repone rate in GATS i preented in Chapter 10 of thi manual a well a in GATS Quality Aurance: Guideline and Documentation. 3. In ome tudie, ample ubtitution are allowed. Thi refer to the practice of replacing eligible nonrepondent or ineligible with another randomly elected unit from the ample frame. Sample ubtitution can lead to biaed etimate and therefore are not allowed with GATS. REQUIREMENT: No ubtitution for eligible nonrepondent or ineligible. 4. Sample upplementation refer to the practice of introducing an additional ample into the urvey. Thi i done to increae the final repondent ample ize when the level of ineligibility and/or nonrepone are higher than anticipated in a particular tratum or area. Sample upplementation i allowed provided there i ufficient time in the data collection chedule to allow the new cae to proceed through all phae of data collection in a typical manner. 5. A maximum of one repondent per houehold hould be elected for an interview. Since handheld computer are ued for houehold rotering and data collection in all GATS data collection operation, the handheld device will ue a random number generator to elect the repondent and diplay the name of the choen individual on the creen. Detail on within-houehold ampling are preented in Chapter 9. If any country implementing the GATS ha quetion regarding the deign requirement and recommendation preented in thi ection, or if a country would like to requet a deviation from them, it hould contact CDC or WHO. Global Adult Tobacco Survey (GATS) 4-6 Sample Deign Manual Chapter 4: Baic Survey Deign Specification

27 5. Overview of the Sampling Approach The ample for GATS hould be elected uing a multi-tage, geographically clutered deign to enure adequate coverage of the entire target population while imultaneouly minimizing data collection cot. Cot are reduced becaue the geographic clutering of the ample deign will minimize the travel needed for interviewer to viit each elected houehold. Thi deign i outlined in Exhibit 5-1. In ummary, thi ample election proce hould proceed a follow: The firt tage() of the deign will involve electing etablihed geo-political area unit. Thi election proce will be highly dependent on the particular country that i fielding GATS. Some countrie may have uitable data o that only one tage of geographic election i needed. Other countrie may need multiple tage of geographic election, where each tage i deigned to randomly ubelect geographic area from within larger area elected at a previou tage. REQUIREMENTS: Select at leat 100 geographic PSU and 400 egment. Select geographic area uing a random (e.g. ytematic) PPS technique. Size i eligible population count (preferred) or total The ultimate goal i to elect at leat 100 primary ampling unit houehold. and at leat 400 ultimate geographic unit in the urvey. Ultimate Select at leat 100 PSU and 400 geographic unit, alo referred to a ample egment, are the egment from each region if mallet geographic area elected for GATS. Note that if a country want to compare etimate country i electing geographic area in only one tage of by urbanicity within the region. election, then we would expect them to ample about 400 Otherwie elect 50 PSU and 200 egment directly from their initial area frame ince the egment i egment per region. actually the primary ampling unit in thi urvey deign. Alo, ome country ample deign may call for intermediate tage of Houehold within egment hould be ampling to elect egment within ample PSU. randomly elected with equal probability and without-replacement. Geographic area hould be elected at each tage with probability proportionate to ize (PPS). The ize meaure can be contructed from a recent cenu and/or adminitrative record and, in general, hould equal the total eligible population, 15 year of age or older, or the number of houehold in the area. If thee count are not available, correponding total population or houehold count can be ued a ize meaure. If regional etimate are deired and a country i intereted in comparing rural area of the region with urban area, then a country hould elect at leat 100 PSU and at leat 400 egment from each region with probability proportionate to ize. If regional etimate are deired but a country i not intereted in comparing rural area of the region with urban area, then at leat 50 primary ampling unit and at leat 200 egment hould be elected from each region with probability proportionate to ize. After the final phae of geographic area ampling i completed (i.e., the egment ampling tage), eparate liting of all houehold that reide in each ample egment hould be produced. Thee liting can be contructed uing adminitrative record, by the field upervior reponible for viiting elected houehold in the area, or by other taff pecially trained in counting and liting houehold in a elected area. An example of the kind of houehold liting that hould be contructed for each elected Global Adult Tobacco Survey (GATS) 5-1 Sample Deign Manual Chapter 5: Overview of the Sampling Approach

28 area, a well a ome recommendation on how to create thi lit during a viit to the area, can be found in the GATS Mapping and Liting Manual. The houehold liting within each ample egment will erve a a ample frame for the election of houehold within each elected egment. Houehold hould be randomly elected from the lit with equal probability and without-replacement. Simple random ampling or ytematic ampling after a random tart can be ued. After the houehold ample i elected in each egment, and if the country ha decided to do gender randomization, each ampled houehold hould be randomly aigned to either the male houehold group or the female houehold group. Houehold in the male group will have only urvey-eligible male rotered during the creening viit and ubequently will have only male elected for the GATS interview. Similarly, houehold in the female group will have only urvey-eligible female rotered during the creening viit and ubequently will have only female elected for the GATS interview. (For more on gender randomization, ee Chapter 2.) Interviewer will viit each elected houehold. If thi viit reveal that additional houehold have emerged in an area or if the interviewer ee that the elected houehold i in fact a multiple houehold tructure (uch a an apartment building), then ome ubampling may be required by the interviewer in conultation with the field upervior. Thi type of ubampling i not likely to be needed very often. After a lited addre i elected and confirmed to be a ingle houehold, the interviewer will viit the houehold and create a roter of all urvey-eligible male and or female (depending on if gender randomization i ued) who would conider the ampled houehold to be their uual place of reidence. (For more on criteria for reidency and memberhip in the GATS target population, ee Section 3.1.) For countrie adminitering the urvey on handheld computer, the handheld device will ue a random number generator to elect one individual from within the houehold. Only thee elected individual will be adminitered the GATS Individual Quetionnaire ince no ubtitution for nonrepone i allowed. A more detailed dicuion of each tage of the ampling proce i provided in the next few ection. Global Adult Tobacco Survey (GATS) 5-2 Sample Deign Manual Chapter 5: Overview of the Sampling Approach