Methods for Categorical Longitudinal Survey Data: Understanding Employment Status of Australian Women

Many variables that are of interest in social science research are nominal variables with two or more categories, such as employment status, occupation, political preference, or self-reported health status. With longitudinal survey data it is possible to analyse the transitions of individuals between different employment states or occupations (for example). In the statistical literature, models for analysing categorical dependent variables with repeated observations belong to the family of models known as generalized linear mixed models (GLMMs). The specific GLMM for a dependent variable with three or more categories is the multinomial logit random effects model. For these models, the marginal distribution of the response does not have a closed form solution and hence numerical integration must be used to obtain maximum likelihood estimates for the model parameters. Techniques for implementing the numerical integration are available but are computationally intensive requiring a large amount of computer processing time that increases with the number of clusters (or individuals) in the data and are not always readily accessible to the practitioner in standard software. For the purposes of analysing categorical response data from a longitudinal social survey, there is clearly a need to evaluate the existing procedures for estimating multinomial logit random effects model in terms of accuracy, efficiency and computing time. The computational time will have significant implications as to the preferred approach by researchers. In this paper we evaluate statistical software procedures that utilise adaptive Gaussian quadrature and MCMC methods, with specific application to modeling employment status of women using a GLMM, over three waves of the HILDA survey.

Measuring globalisation: An evolutionary economic approach to tracking the evolution of international trade

Operationalising and measuring the concept of globalisation is important, as the extent to which the international economy is integrated has a direct impact on industrial dynamics, national trade policies and firm strategies. Using complex systems network analysis with longitudinal trade data from 1938 to 2003, this paper presents a new way to measure globalisation. It demonstrates that some important aspects of the international trade network have been remarkably stable over this period. However, several network measures have changed substantially over the same time frame. Taken together, these analyses provide a novel measure of globalisation.

Multi-level zero-inflated Poisson regression modelling of correlated count data with excess zeros

Count data with excess zeros relative to a Poisson distribution are common in many biomedical applications. A popular approach to the analysis of such data is to use a zero-inflated Poisson (ZIP) regression model. Often, because of the hierarchical Study design or the data collection procedure, zero-inflation and lack of independence may occur simultaneously, which tender the standard ZIP model inadequate. To account for the preponderance of zero counts and the inherent correlation of observations, a class of multi-level ZIP regression model with random effects is presented. Model fitting is facilitated using an expectation-maximization algorithm, whereas variance components are estimated via residual maximum likelihood estimating equations. A score test for zero-inflation is also presented. The multi-level ZIP model is then generalized to cope with a more complex correlation structure. Application to the analysis of correlated count data from a longitudinal infant feeding study illustrates the usefulness of the approach.

A Bayesian hierarchical model for categorical longitudinal data from a social survey of immigrants

The paper investigates a Bayesian hierarchical model for the analysis of categorical longitudinal data from a large social survey of immigrants to Australia. Data for each subject are observed on three separate occasions, or waves, of the survey. One of the features of the data set is that observations for some variables are missing for at least one wave. A model for the employment status of immigrants is developed by introducing, at the first stage of a hierarchical model, a multinomial model for the response and then subsequent terms are introduced to explain wave and subject effects. To estimate the model, we use the Gibbs sampler, which allows missing data for both the response and the explanatory variables to be imputed at each iteration of the algorithm, given some appropriate prior distributions. After accounting for significant covariate effects in the model, results show that the relative probability of remaining unemployed diminished with time following arrival in Australia.

Do the Genetic or Environmental Determinants of Anxiety and Depression Change with Age? A Longitudinal Study of Australian Twins

Because the determinants of anxiety and depression in late adolescence and early adulthood may differ from those in later life, we investigated the temporal stability and magnitude of genetic and environmental correlates of symptoms of anxiety and depression across the life span. Data were collected from a population-based Australian sample of 4364 complete twin pairs and 777 singletons aged 20 to 96 years who were followed-up over three studies between 1980 and 1996. Each study contained the 14-item self-report DSSI/sAD scale which was used to measure recently experienced symptoms of anxiety and depression. Symptom scores were then divided and assigned to age intervals according to each subject's age at time of participation. We fitted genetic simplex models to take into account the longitudinal nature of the data. For male anxiety and depression, the best fitting simplex models comprised a single genetic innovation at age 20 which was transmitted, and explained genetic variation in anxiety and depression at ages 30, 40, 50 and 60. Most of the lifetime genetic variation in female anxiety and depression could also be explained by innovations at age 20 which were transmitted to all other ages; however, there were also smaller age-dependent genetic innovations at 30 for anxiety and at 40 and 70 for depression. Although the genetic determinants of anxiety and depression appear relatively stable across the life-span for males and females, there is some evidence to support additional mid-life and late age gene action in females for depression. The fact that mid-life onset for anxiety occurs one decade before depression is also consistent with a causal relationship (anxiety leading to depression) between these conditions. These findings have significance for large scale depression prevention projects.

Longitudinal relationships among overweight, life satisfaction, and aspirations in young women

Objective: This study investigated associations of overweight status and changes in overweight status over time with life satisfaction and future aspirations among a community sample of young women. Research Methods and Procedures: A total of 7865 young women, initially 18 to 23 years of age, completed two surveys that were 4 years apart. These women provided data on their future life aspirations in the areas of further education, work/career, marital status, and children, as well as their satisfaction with achievements to date in a number of life domains. Women reported their height and weight and their sociodemographic characteristics, including current socioeconomic status (occupation). Results: Young women's aspirations were cross-sectionally related to BMI category, such that obese women were less likely to aspire to further education, although this relationship seemed explained largely by current occupation. Even after adjusting for current occupation, young women who were obese were more dissatisfied with work/career/study, family relationships, partner relationships, and social activities. Weight status was also longitudinally associated with aspirations and life satisfaction. Women who were overweight or obese at both surveys were more likely than other women to aspire to other types of employment (including self-employed and unpaid work in the home) as opposed to full-time employment. They were also less likely to be satisfied with study or partner relationships. Women who resolved their overweight/obesity status were more likely to aspire to being childless than other women. Discussion: These results suggest that being overweight/obese may have a lasting effect on young women's life satisfaction and their future life aspirations.

Country of birth, country of residence, and menopausal transitions and symptoms: British birth cohort and Australian Longitudinal Study on Women's Health

Objective: To explore endocrine-related and general symptoms among three groups of middle-aged women defined by country of birth and country of residence, in the context of debates about biological, cultural and other factors in menopause. Methods: British-born women participating in a British birth cohort study (n=1,362) and age-matched Australian-born (n=1,724) and British-born (n=233) Australian women selected from the Australian Longitudinal Study on Women's Health (ALSWH) responded to two waves of surveys at ages 48 and 50. Results: Australian-Australian and British-Australian women report reaching menopause later than British-British women, even after accounting for smoking status and parity. Hormone replacement therapy (HRT) use was lower and hysterectomy was more common among both Australian groups, probably reflecting differences in health services between Britain and Australia. The Australian-Australian and British-Australian groups were more likely to report endocrine-related symptoms than the British-British group, even after adjusting for menopausal status. British-British women were more likely to report some general symptoms. Conclusions: Symptom reporting is high among Australian and British midlife women and varies by country of residence, country of birth and menopausal status. Implications: The data do not support either a simple cultural or a simple biological explanation for differences in menopause experience.

Multiple imputation for body mass index: lessons from the Australian Longitudinal Study on Women's Health

In large epidemiological studies missing data can be a problem, especially if information is sought on a sensitive topic or when a composite measure is calculated from several variables each affected by missing values. Multiple imputation is the method of choice for 'filling in' missing data based on associations among variables. Using an example about body mass index from the Australian Longitudinal Study on Women's Health, we identify a subset of variables that are particularly useful for imputing values for the target variables. Then we illustrate two uses of multiple imputation. The first is to examine and correct for bias when data are not missing completely at random. The second is to impute missing values for an important covariate; in this case omission from the imputation process of variables to be used in the analysis may introduce bias. We conclude with several recommendations for handling issues of missing data. Copyright (C) 2004 John Wiley Sons, Ltd.

The impact of breast cancer on the lives of middle-aged women: Results from the Australian longitudinal study of women's health

This article investigated the impact of breast cancer (BC) in middle-aged Australian women (45-50 years). Two waves of data collected 2 years apart from a longitudinal survey of 12,177 women identified 3 groups: (a) 11,933 (98%) who reported never having had BC, (b) 181 (1.5%) who reported a diagnosis of BC at Time 1, and (c) 63 (0.5%) who reported onset of BC between Time I and Time 2. Repeated measures analysis of variance was used to compare the 3 groups. Women with recent onset of BC experienced significant changes across a range of functioning compared with the other 2 groups. Compared with women with no BC, women with longer established onset of BC had significantly worse health and social outcomes, but these were associated with small effect sizes. Both groups of women with BC reported less impact on mental and emotional health than on other areas of functioning.

Asthma and internalizing behavior problems in adolescence: A longitudinal study

Objective: Clinical studies of asthmatic children have found an association between lung disease and internalizing behavior problems. The causal direction of this association is, however, unclear. This article examines the nature of the relationship between behavior and asthma problems in childhood and adolescence. Methods: Data were analyzed on 5135 children from the Mater University Study of Pregnancy and its outcomes (MUSP), a large birth cohort of mothers and children started in Brisbane, Australia, in 1981. Lung disease was measured from maternal reports of asthma/bronchitis when the children were aged 5 and maternal reports of asthma symptoms when the children were aged 14. Symptoms of internalizing behaviors were obtained by maternal reports (Child Behavior Checklist) at 5 years and by maternal and children's reports at 14 years (Child Behavior Checklist and Youth Self Report). Results: Although there was no association between prevalence of asthma and externalizing symptoms, asthma and internalizing symptoms were significantly associated in cross-sectional analyses at 5 and 14 years. In prospective analyses, after excluding children with asthma at 5 years, internalizing symptoms at age 5 were not associated with the development of asthma symptoms at age 14. After excluding children with internalizing symptoms at 5 years, those who had asthma at 5 years had greater odds of developing internalizing symptoms at age 14. Conclusion: Children who have asthma/bronchitis by the age of 5 are at greater risk of having internalizing behavior problems in adolescence.

Does birth weight predict childhood diet in the Avon longitudinal study of parents and children?

Study objective: Low birth weight predicts cardiovascular disease in adulthood, and one possible explanation is that children with lower birth weight consume more fat than those born heavier. Therefore, the objective of this study was to investigate associations between birth weight and childhood diet, and in particular, to test the hypothesis that birth weight is inversely related to total and saturated fat intake. Design: Prospective cohort study. Setting: South west England. Participants: A subgroup of children enrolled in the Avon longitudinal study of parents and children, with data on birth weight and also diet at ages 8, 18, 43 months, and 7 years ( 1152, 998, 848, and 771 children respectively). Main results: Associations between birth weight and diet increased in strength from age 8 to 43 months, but had diminished by age 7 years. Fat, saturated fat, and protein intakes were inversely, and carbohydrate intake was positively associated with birth weight at 43 months of age, after adjusting for age, sex, and energy intake. After adjustment for other confounders, all associations were weakened, although there was still a suggestion of a relation with saturated fat ( -0.48 (95% CI -0.97, 0.02) g/day per 500 g increase in birth weight. Similar patterns were seen in boys and girls separately, and when the sample was restricted to those with complete data at all ages. Conclusions: A small inverse association was found between birth weight and saturated fat intake in children at 43 months of age but this was not present at 7 years of age. This study therefore provides little evidence that birth weight modifies subsequent childhood diet.

A Comparison of Methodologies from Two Longitudinal Community-Based Randomized Controlled Trials of Similar Interventions in Palliative Care: What Worked and What Did Not

Background: Methodological challenges such as recruitment problems and participant burden make clinical trials in palliative care difficult. In 2001-2004, two community-based randomized controlled trials (RCTs) of case conferences in palliative care settings were independently conducted in Australia-the Queensland Case Conferences trial (QCC) and the Palliative Care Trial (PCT). Design: A structured comparative study of the QCC and PCT was conducted, organized by known practical and organizational barriers to clinical trials in palliative care. Results: Differences in funding dictated study designs and recruitment success; PCT had 6 times the budget of QCC. Sample size attainment. Only PCT achieved the sample size goal. QCC focused on reducing attrition through gatekeeping while PCT maximized participation through detailed recruitment strategies and planned for significant attrition. Testing sustainable interventions. QCC achieved a higher percentage of planned case conferences; the QCC strategy required minimal extra work for clinicians while PCT superimposed conferences on normal work schedules. Minimizing participant burden. Differing strategies of data collection were implemented to reduce participant burden. QCC had short survey instruments. PCT incorporated all data collection into normal clinical nursing encounters. Other. Both studies had acceptable withdrawal rates. Intention-to-treat analyses are planned. Both studies included substudies to validate new outcome measures. Conclusions: Health service interventions in palliative care can be studied using RCTs. Detailed comparative information of strategies, successes and challenges can inform the design of future trials. Key lessons include adequate funding, recruitment focus, sustainable interventions, and mechanisms to minimize participant burden.