James R. Abernathy

Application of the Randomized Response Technique in Obtaining Quantitative Data

Abstract The randomized response technique of reducing respondent bias in obtaining answers to sensitive questions is extended from the situation where response is categorical to that in which the response is quantitative. Results are reported on the application of the method to estimating mean number of abortions in an urban population of women, and mean income of heads of households. The efficiency of estimators based on the method of moments in the randomized response procedure is studied and representative results are reported and discussed.

Estimates of induced abortion in urban North Carolina

In 1965, Warner developed an interviewing procedure designed to eliminate evasive answer bias when questions of a sensitive nature are asked. He called the procedure ‘randomized response.’ The authors have been studying the technique for several years and, in this paper, are reporting some of the estimates of induced abortion in urban North Carolina using randomized response. Estimates of the proportion of women having an abortion during the past year among women 18–44 years of age are reported. For the study population indices were developed relating induced abortion to total conceptions for whites and nonwhites. The illegal abortion rate per 100 conceptions was estimated to be 14.9 for whites and 32.9 for nonwhites. Estimates of the proportion of women having an abortion during their lifetime among women 18 years old or over are also shown. Among ever married women, the proportion having an abortion during their lifetime declined as education increased. Estimates were high for women with 5 or more pregnancies. Most of the respondents stated that they were satisfied that the randomized response approach would not reveal their personal situation. Furthermore, they did not think their friends would truthfully respond to adirect question regarding abortion.

Association of Alcohol Consumption to Mortality in Middle-Aged U.S. and Russian Men and Women

PURPOSE The purpose of this paper is to assess the relationships of alcohol consumption to total and cardiovascular disease mortality in U.S. and Russian men and women after adjustment for several covariates. A secondary objective is to determine how this relationship varies by country and gender. METHODS Men aged 40-59 and women aged 40-69 screened in Russia and the U.S. between 1972 and 1982 were followed for mortality for 13 years as part of the Lipid Research Clinics Prevalence and Follow-up Studies. Alcohol consumption was based on a 7-day recall of drinks of beer, wine, mixed drinks, and liquors. Drinkers of alcohol were classified into four levels based on amount consumed during the recall period. RESULTS Age-adjusted mortality rates were higher for non-drinkers than lower level drinkers in both genders and countries, and there was an indication that mortality rates for high level drinkers, especially in men, approached those of non-drinkers. When mortality rates were adjusted for other risk factors they remained higher for non-drinkers in U.S. men and women, but in Russia, with one exception, there was no difference in mortality rates between drinkers and non-drinkers. Relative risks for cardiovascular disease mortality rates were similar to those of total mortality. CONCLUSIONS Beneficial association of alcohol consumption and mortality may be limited depending upon the prevalence of other risk factors in the studied population.

Risk Assessment for Oral Diseases

This paper seeks to achieve four goals, each of which forms the basis for a section in the presentation. First, the rationale of risk assessment is fully described. In this section, some of the necessary conditions are identified that make disease prediction worth pursuing. The second section discusses some essential background to the understanding of risk assessment in dentistry. In this segment, attention is focused on population-based and individual-based perspectives, alternative approaches to expressing health risk, and methods for comparing the predictive accuracy of alternative risk assessment models. The third section of the paper develops a conceptual frameworkforrisk assessment in dentistry. Particular emphasis is devoted to the identification of risk factors and their incorporation into alternative statistical models. In the fourth section, empirical data are offered by which certain comparisons of the alternative risk models can be drawn. The paper concludes with a discussion that emphasizes data and technical limitations, speculates on future applications, and suggests new avenues for research.

Association of high-density-lipoprotein cholesterol with mortality and other risk factors for major chronic noncommunicable diseases in samples of US and Russian men

Previous reports from the Russian Lipid Research Clinics (LRC) study showed no association between the level of high-density-lipoprotein (HDL) cholesterol and mortality from coronary heart disease (CHD), while US LRC data indicated a strong negative association between HDL cholesterol and CHD mortality. This report investigated the association of HDL cholesterol and mortality in these same population samples with follow-up extended to 12 years. The association between HDL cholesterol and mortality remained inverse and significant in the US sample. In the Russian sample, high levels of HDL cholesterol were associated with higher risk of all-cause and cancer mortality, although adjustment for known risk factors reduced the strength of the association. The association between HDL cholesterol and CHD mortality was negative in the Russian sample, although the strength of the association was less than that for the US sample. Extended follow-up reduced the difference in the association between HDL cholesterol and mortality between the two countries; however, important differences remained. Further research will be required to clearly determine the cause for their differences.

Respondent hazards in the unrelated question randomized response model

Abstract The randomized response technique is a method of interviewing which allows estimation of population proportion or mean of a sensitive characteristic while maintaining anonymity of response. This is accomplished by use of two questions in the interview, one sensitive and one not so, with the respondent applying a random device to determine which question to answer. The statistician knows the parameters of the random device but is kept from knowledge of individual outcomes. When the nonsensitive question bears on a matter quite foreign to the sensitive characteristic, it is termed an unrelated question, and the associated technique carries that designation. Critical in the construction of such a model are the assigned probability p that the respondent will select the sensitive question and the proportion Πy of the population having the unrelated nonsensitive characteristic. The choice of these parameters is discussed herein with reference to the risks perceived by respondents.

Application of discriminant functions in perinatal death and survival

Abstract Three discriminant functions were used to determine which were the most effective in discriminating between death and survival of an infant, and to assess the importance of each independent variable in the analyses. Model A contained 28 variables which had been found significant or borderline in previous regression analyses upon birth weight, gestation, or crown-heel length. Model B consisted only of birth weight, gestation, and crown-heel length, while Model C contained the 25 variables of Model A not included in Model B. Model A was judged to be the most effective in discrimination. Model B, with only three variables, was almost as powerful. Model C, with 25 variables, was the least effective by far. The three variables of Model B were further analyzed to determine which of these three indices of prematurity, and which combinations of them were most important in discriminating between death and survival. Birth weight was the most powerful, gestation was next, and crown-heel length the least important. Considering the variables in pairs, birth weight and gestation was the most effective pair, followed by birth weight and crown-heel length. Gestation and crown-heel length was the least effective. Birth weight alone was a better discriminator than gestation and length combined. Furthermore, the addition of a third index to any given pair significantly increased the discriminating power of the function. Individual variables, other than the three discussed above, found significant in discrimination included race, sex, prepregnancy weight of the mother, obstetric complications, placenta and cord conditions, and congenital malformations.

On A Method for Studying Family Size Preferences

The concept of desired family size is widely accepted, and numerous studies have been undertaken to measure its level in certain population groups. The determination of desired family size has generally followed the direct question procedure through which the respondent states some whole number as representing her family size wishes at some specified period in her life. Limitations of this method have been recognized, and some research has been conducted toward the development of other techniques of measuring desired family size.The purpose of this paper is to discuss a method of estimating desired family size based on the concept of paired comparison, where the response is restricted in order to reduce respondent bias. Emphasis is on estimation of parameters and test statistics appropriate for evaluating the applicability of the underlying model.

Fetal, parental, and environmental factors associated with perinatal mortality in mothers under 20 years of age.

Abstract 1. 1. Excessive parity in young mothers and illegitimacy are associated with higher perinatal mortality rate. 2. 2. This increased perinatal mortality rate involves both fetal and neonatal components with the neonatal component predominant. 3. 3. Prematurity and toxemia of pregnancy are largely responsible for the increased mortality observed among women under 20 years of age. 4. 4. The socioeconomic status of the patient in this age group is extremely poor and probably accounts for the increased incidence of prematurity and toxemia which are observed.

Blood pressure and heart rate response during exercise in men and women in the USA and Russia lipid research clinics prevalence study

Examination of blood pressure (BP) and heart rate (HR) measurements at rest and during exercise in samples of USA and Russian middle-aged men and women show significant differences between countries for both genders. Russian men had higher resting systolic blood pressure (SBP) than USA men but lower SBP at both stages of exercise. Russian women had significantly higher resting SBP than USA women at rest and also during exercise. Russian men and women had significantly lower HR at rest and during exercise than USA men and women. Differences between countries were also noted for mean body weight, height, lipid levels, percentages of smokers and several other variables, and it was postulated that differences in these factors might be responsible in part for differences in SBP and HR during exercise. After adjustment for these variables, differences in SBP during exercise between USA and Russian men remained, but differences between USA and Russian women disappeared. For both genders, differences in HR during exercise remained after adjustment. Mortality analyses in USA and Russian men indicated that stage 2 SBP response during exercise was generally not a significant mortality risk factor after adjustment for age, BMI, TC, smoking and resting SBP. In a similar model, stage 2 HR response during exercise was also a non-significant risk factor in mortality.