This file includes Stata codes and all data required for replicating results in the article "Trends in Mail-Order Pharmacy Use in the U.S. From 1996 to 2018: Analysis of the Medical Expenditure Panel Survey" published at the American Journal of Preventive Medicine.

Background: Increasing numbers of variables in surveys and administrative databases are created. Principal component analysis (PCA) is important to summarize data or reduce dimensionality. However, one disadvantage of using PCA is the interpretability of the principal components (PCs), especially in a high-dimensional database. By analyzing the variance distribution according to PCA loadings and approximating PCs with input variables, we aim to demonstrate the importance of variables based on the proportions of total variances contributed or explained by input variables.Methods: There were five data sets of various sizes used to understand the performance of PC approximation: Hitters, SF-12v2 subset of the 2004–2011 Medical Expenditure Panel Survey (MEPS), and the full set of 1996–2011 MEPS data, along with two data sets derived from the Canadian Health Measures Survey (CHMS): a spirometry subset with the measures from the first trial of spirometry and a full data set that contained non-redundant variables. The variables in data sets were first centered and scaled before PCA. PCs were approximated through two approaches. First, the PC loadings were squared to estimate the variance contribution by variables to PCs. The other method was to use forward-stepwise regression to approximate PCs with all input variables.Results: The first few PCs had large variances in each data set. Approximating PCs using stepwise regression could efficiently identify the input variables that explain large portions of PC variances than approximating according to PCA loadings in the data sets. It required fewer numbers of variables to explain more than 80% of the PC variances through stepwise regression.Conclusion: Approximating and interpreting PCs with stepwise regression is highly feasible.PC approximation is useful to (1) interpret PCs with input variables, (2) understand the major sources of variances in data sets, (3) select unique sources of information, and (4) search and rank input variables according to the proportions of PC variance explained. This can be an approach to systematically understand databases and search for variables that are important to databases.

This longitudinal survey was designed to add significantly to the amount of detailed information available on the economic situation of households and persons in the United States. These data examine the level of economic well-being of the population and also provide information on how economic situations relate to the demographic and social characteristics of individuals. There are three basic elements contained in the survey. The first is a control card that records basic social and demographic characteristics for each person in a household, as well as changes in such characteristics over the course of the interviewing period. The second element is the core portion of the questionnaire, with questions repeated at each interview on labor force activity, types and amounts of income, participation in various cash and noncash benefit programs, attendance in postsecondary schools, private health insurance coverage, public or subsidized rental housing, low-income energy assistance, and school breakfast and lunch participation. The third element consists of topical modules which are series of supplemental questions asked during selected household visits. No topical modules were created for the first or second waves. The Wave III Rectangular Core and Topical Module File offers both the core data and additional data on (1) education and work history and (2) health and disability. In the areas of education and work history, data are supplied on the highest level of schooling attained, courses or programs studied in high school and after high school, whether the respondent received job training, and if so, for how long and under what program (e.g., CETA or WIN). Other items pertain to the respondent's general job history and include a description of selected previous jobs, duration of jobs, and reasons for periods spent not working. Health and disability variables present information on the general condition of the respondent's health, functional limitations, work disability, and the need for personal assistance. Data are also provided on hospital stays or periods of illness, health facilities used, and whether health insurance plans (private or Medicare) were available. Respondents whose children had physical, mental, or emotional problems were questioned about the causes of the problems and whether the children attended regular schools. The Wave IV Rectangular Core and Topical Module file contains both the core data and sets of questions exploring the subjects of (1) assets and liabilities, (2) retirement and pension coverage, and (3) housing costs, conditions, and energy usage. Some of the major assets for which data are provided are savings accounts, stocks, mutual funds, bonds, Keogh and IRA accounts, home equity, life insurance, rental property, and motor vehicles. Data on unsecured liabilities such as loans, credit cards, and medical bills also are included. Retirement and pension information covers such items as when respondents expect to stop working, whether they will receive retirement benefits, whether their employers have retirement plans, if so whether they are eligible, and how much they expect to receive per year from these plans. In the category of housing costs, conditions, and energy usage, variables pertain to mortgage payments, real estate taxes, fire insurance, principal owed, when the mortgage was obtained, interest rates, rent, type of fuel used, heating facilities, appliances, and vehicles. The Wave V topical modules explore the subject areas of (1) child care, (2) welfare history and child support, (3) reasons for not working/reservation wage, and (4) support for nonhousehold members/work-related expenses. Data on child care include items on child care arrangements such as who provides the care, the number of hours of care per week, where the care is provided, and the cost. Questions in the areas of welfare history and child support focus on receipt of aid from specific welfare programs and child support agreements and their fulfillment. The reasons for not working/reservation wage module presents data on why persons are not in the labor force and the conditions under which they might join the labor force. Additional variables cover job search activities, pay rate required, and reason for refusal of a job offer. The set of questions dealing with nonhousehold members/work-related expenses contains items on regular support payments for nonhousehold members and expenses associated with a job such as union dues, licenses, permits, special tools, uniforms, or travel expenses. Information is supplied in the Wave VII Topical Module file on (1) assets and liabilities, (2) pension plan coverage, and (3) real estate property and vehicles. Variables pertaining to assets and liabilities are similar to those contained in the topical module for Wave IV. Pension plan coverage items include whether the respondent will receive retirement benefits, whether the employer offers a retirement plan and if the respondent is included in the plan, and contributions by the employer and the employee to the plan. Real estate property and vehicles data include information on mortgages held, amount of principal still owed and current interest rate on mortgages, rental and vacation properties owned, and various items pertaining to vehicles belonging to the household. Wave VIII Topical Module includes questions on support for nonhousehold members, work-related expenses, marital history, migration history, fertility history, and household relationships. Support for nonhousehold members includes data for children and adults not in the household. Weekly and annual work-related expenses are documented. Widowhood, divorce, separation, and marriage dates are part of the marital history. Birth expectations as well as dates of birth for all the householder's children, in the household or elsewhere, are recorded in the fertility history. Migration history data supplies information on birth history of the householder's parents, number of times moved, and moving expenses. Household relationships lists the exact relationships among persons living in the household. Part 49, Wave IX Rectangular Core and Topical Module Research File, includes data on annual income, retirement accounts, taxes, school enrollment, and financing. This topical module research file has not been edited nor imputed, but has been topcoded or bottomcoded and recoded if necessary by the Census Bureau to avoid disclosure of individual respondents' identities. (Source: downloaded from ICPSR 7/13/10)

Please Note: This dataset is part of the historical CISER Data Archive Collection and is also available at ICPSR at https://doi.org/10.3886/ICPSR08317.v2. We highly recommend using the ICPSR version as they may make this dataset available in multiple data formats in the future.

Abstract

At the time of conception of VPA-2, the main purpose of the survey was to provide the baseline for the next National Development Plan. Equally important, it was to become the main source for the Maldives' first initiative in MDG tracking and reporting. Finally, as it included a 'panel' survey (same households, with similar questions seven years later), it could indicate successful coping mechanisms and poverty reduction strategies at the household level.

The survey acquired even greater significance as a result of the tsunami on 26th December 2004. The fieldwork that had been completed in July 2004 provided a detailed description of the socioeconomic conditions on the islands only a few months prior to the disaster. The Government was thus able to use the comparative analysis, which was already under way, to make preliminary estimates of the effects of the tsunami on people's livelihoods in the affected islands.

More specifically, VPA-2 aimed to include:

1. The basis for an anti-poverty framework - An in-depth analysis of living conditions in all parts of the country should form the basis for a strategic anti-poverty framework. This should enable the Government to design pro-poor policies and programmes, as well as monitor and evaluate their impact.

2. The people's perspective - The VPA was to provide an assessment, both in terms of geographical coverage and the range of development concerns, of the needs and priorities from the perspective of the people themselves. This was to include a human vulnerability index (HVI) tailored for a scattered and extensive island state.

3. A database - Provide a relational database for poverty and vulnerability diagnostics;

4. An evaluation - Looking at the effects of development activities upon household living standards.

The VPA-2 would then serve as the cornerstone for actions in a number of areas, including:

1. Millennium Development Goals - A analysis of Millennium Development Goal (MDG) indicators and the writing of the first Maldives MDG Report;

2. Public finance - A discussion of the allocative aspect of public finance and budgeting and social spending, arising from the results of the World Bank public expenditure report;

3. Development plans - Data support for an evaluation of the current Sixth National Development Plan (NDP) and the formulation of the Seventh NDP.

The Government's decision to embark on this exercise reflects the importance it attached to the availability of comprehensive socioeconomic data for policy formulation. VPA-2 would not only highlight continuing problems, but also assess the effects of government policies. The panel data in particular would provide a sample large enough to allow for an in-depth analysis of changes in poverty and living conditions of households across the nation - and indicate why some households had made more progress than others.

Geographic coverage

National, Male', Atolls

Analysis unit

• Households
• Individuals
• Children under 5 years
• Consumption expenditure items/ services
• Community (to supplement the household information)

Universe

The survey covered all households including their members. Institutions like hospitals, clinics, hostels, hotels, jails, labour quarters and defence force camps have not been included in the scope of the survey. However, staff members of the above mentioned institutions living independently in premises attached to these institutions have been included.

Kind of data

Sample survey data [ssd]

Sampling procedure

SAMPLING FRAME

The main database for the frame of VPA sampling comes from the Population Census 2000.

In the case of atolls, the frame for the VPA is the list of inhabited islands. The list is most recent and accurate since all inhabited islands were surveyed. The Census 2000 reported 200 inhabited islands. Sampling of islands was not considered appropriate because the level of vulnerability is very much determined by the local conditions.

The frame for Male' consists of 5 wards and 317 enumeration blocks which were created in the last census. These blocks served as the primary sampling units in VPA sample design for Male'.

LEVEL OF REPRESENTATION

The survey has two domains, Male' and Atolls and each of these domains have independent sampling schemes.

SAMPLING IN ATOLLS

The survey covered all 200 inhabited islands where the islands virtually served as independent strata. Since population sizes differ across the strata, the following rule was used in the selection of households: - for islands with 1500 or less inhabitants (approximately 200 households), a minimum of 10 households were allocated for each island - for islands with more than 1500 inhabitants , the sampling rate was increased by 10 households for every 1500 inhabitants

Partial Overlapping Sample

In order to ensure the data comparability of two surveys, half of the samples in all islands were retained from those selected for VPA 97. Some of the advantages of partial overlapping samples for successive surveys include:

1. It balances the advantages and disadvantages of a completely repeated panel and taking independent samples in the successive period. The former can give information about the changes of variables of interest, but ignores the effect of changes outside the panel. The latter scheme, on the other hands, cannot measure the changes occurred in individual units.

2. By using the same sampling units in the successive survey, there are certain gains in the reduction of the variance since the high degree of correlation between the samples of the periods increases the value of the correlation coefficient, thereby reducing the variance.

In obtaining partial overlapping samples for VPA 2004, the small islands each had 5 new households while for the larger islands, half of the households in each stratum were new households.

A total of 2840 households were sampled from atolls in VPA 2004.

Selection Procedure

Selection of households was done using systematic sampling with a random start. Ideally, the list should have been arranged in a systematic manner with a fixed route (clockwise or counter clockwise) so that samples taken from this ordered list creates implicit strata of each interval.

Replacement Scheme for Panel Households

At first, households in the panel of VPA 97 should be identified in the new list. If all households are found, sampling procedure begins. In cases where "old households" could not be found, different rules of replacement were applied to different scenarios. Details of these rules are found in Technical Note 3: Sampling Design of the Vulnerability and Poverty Assessment 2004 Report.

SAMPLING IN MALE'

Male' has no panel data, i.e., a completely new set of samples was taken in the island.

Selection Procedure

A two-stage self-weighting design was applied. Male' was stratified into 5 wards and selection was made within each ward. At the first stage, enumeration blocks were selected with probability proportional to the size (PPS) of blocks in terms of the number of households. In the second stage, a fixed number of 10 households were using systematic sampling from each selected block. In such a case, the blocks served as the primary sampling units (PSU) while the households, the secondary sampling units (SSU) or simply elements.

A technical document entitled Vulnerability and poverty assessment survey - 2004 Sampling Design discusses in detail the Sampling Design of VPA 2004. It can also be found in Technical Note 3: Sampling Design of the Vulnerability and Poverty Assessment 2004 Report.

Note: Detailed sample design information is provided in the technical document which is presented in this documentation.

Mode of data collection

Face-to-face [f2f]

Research instrument

Ten questionnaire forms were prepared for VPA 2004 (VPA II). These include:

Form 1. Listing form Form 2. Structure form Form 3. Individual Information, Education, Employment, Nuptiality and Fertility Form 4. Household Form: Living Condition of the Household, Electricity, Availability of Drinking Water, Garbage Disposal, Health Services, Immigration, Crisis, Hardship, Consumer Durables, Travel Abroad, Problems in our Lives, Investment, Perception of Economic Status and Well Being, Voluntary Work of Household Members, Morbidity, Property Transaction, Loans and Credits Form 5. Measurements of Children under 5 Years Form 6. Employment and Income Form 7. Expenditure Diary: Food purchased in bulk, other food items; Locally produced goods (bought, own produced good) and fresh produce; Fish and fish products; Tobacco / chewing products; Furnishing and furnitures, household items; Clothing and footwear; Housing, water, electricity, gas and fuels; Medical and health expenses; Transport and communication; Education; Entertainment and sports; Personal goods/ personal care; Miscellaneous goods and services Form 8. Crisis and Coping Mechanism Form 9. Problems in our Daily Lives Form 10. Island Form

Cleaning operations

Consistency and plausibility checks were done in the following stages:

1. During data entry: a large number of items were checked for consistency and plausibility. If this process suggested errors, the data entry operators were prompted to cross-check the information they had entered with that on the forms – reducing the number of data transcription errors to an acceptable level while allowing obvious errors to be corrected at an early stage. Once all the data had been entered, more checks for consistency and errors were carried out until an acceptable level of accuracy was obtained and only limited data gaps remained. This was an iterative process demanding frequent