Goodness-of-fit Testing Of Log-link Models For Categorical Outcome Data
Funder
National Health and Medical Research Council
Funding Amount
$260,863.00
Summary
Information about the health consequences of exposure to causal factors is obtained from mathematical models of observed data. Incorrect inferences are possible if the model does not adequately represent the data. Relative risk models are recommended for observations over time on a cohort of subjects, but it is not known how best to assess the adequacy of such models. This project will assess the performance of summary measures of goodness-of-fit when applied to relative risk models.
Novel Statistical Methods For The Analysis Of Meausred Genetic And Environmental Risk Factors In Twin Studies
Funder
National Health and Medical Research Council
Funding Amount
$478,314.00
Summary
Studies on twins are an important way to determine whether the risk of disease is likely to be influenced by genetic factors but have traditionally focussed on unmeasured factors. New epidemiological studies measure thousands of genetic variants on many participants. This project will extend methods for analysing data within and between twin pairs to determine whether risk factors are likely to be causal and therefore should be the subject of further designed studies based on intervention.
Statistical Methods For Handling Missing Data In Longitudinal Studies
Funder
National Health and Medical Research Council
Funding Amount
$198,000.00
Summary
Modern epidemiological research has a strong focus on studying the causes and consequences of major health outcomes over the life span. Studies are increasingly conducted on large cohorts of individuals over long periods of time, extending from before birth through to the later years of life. An example of this type of study is the Victorian Adolescent Health Cohort Study, which began in 1992 with participants aged 15 and is now seeking funding for a 9th wave of data collection in 2005. A major ....Modern epidemiological research has a strong focus on studying the causes and consequences of major health outcomes over the life span. Studies are increasingly conducted on large cohorts of individuals over long periods of time, extending from before birth through to the later years of life. An example of this type of study is the Victorian Adolescent Health Cohort Study, which began in 1992 with participants aged 15 and is now seeking funding for a 9th wave of data collection in 2005. A major challenge that arises in analysing data from studies of this kind is the difficulty created by the occurrence of missing data. In longitudinal studies with multiple measurement occasions, participants rarely complete all waves of data collection, and even when present an individual may not provide data on all study variables. Common practice in analysing such data is to omit individuals entirely if they have a missing value on any of the variables required for the analysis in question. This approach can lead to major biases in conclusions, by excluding individuals in whom patterns of association may be quite different than among those retained, and at best leads to loss of reliability in findings due to the reduction in numbers available for analysis. Recent statistical research has led to a range of new techniques for better handling of missing data in such studies, including the method of multiple imputation (MI), under which multiple copies of the dataset are created with imputed values filled in for the missing values. This approach has enormous potential for helping to produce better answers from large longitudinal studies but a number of issues require research to ensure that the method is made available to researchers in a convenient form and, most importantly, used in a way that leads to sound conclusions. This project will address many of these issues, leading to enhanced capacity to extract valuable information from large epidemiological studies.Read moreRead less