Methodological challenges in public health research: evidence from epidemiological and clinical studies
Population Health Methods and Innovation research theme group, School of Public Health, The University of Queensland
Four members of this theme group will give brief presentations that showcase the diverse nature of methodologies used in public health, epidemiology and clinical research.
Rob Ware - Calculating population attributable fraction when the condition is common
The population attributable fraction (PAF) is the difference in risk of a condition between an exposed population and an unexposed population. It is common to calculate the PAF using logistic regression, however when the condition is common this method is not accurate. We will consider alternative methods to calculate the PAF using the example of respiratory viruses and acute respiratory infection in infants.
Mark Jones - Displaying information on adverse events using data from clinical study reports
In publications of clinical trials, information reported on adverse events is often based on descriptive reporting of incidences of common adverse events with some publications providing no data but rather a sentence stating the intervention was well tolerated. Clinical study reports are detailed reports of clinical trials produced for regulators in applications for drug approval that run to 100s or even 1000s of pages. Detailed information on all coded adverse events is provided for each individual patient in the study. In this short presentation, Mark will summarise ongoing research on how this more detailed information can be displayed graphically to allow a more comprehensive assessment of safety of pharmaceutical interventions.
Nargess Saiepour - Multiple Imputation to skewed data
Multiple imputation is one of the mostly used imputation methods to impute missing data and adjust for attrition bias. This method is mainly used under the mechanism missing at random data and it has important statistical properties. We will compare different approaches in applying multiple imputation to skewed distributions.
Peter Baker - Efficient data analysis using R: a DRY approach from the Australian outback
As a statistical consultant, we often find ourselves repeating the same steps when analysing data for different projects. Reusing R syntax and developing custom R functions helps to improve efficiency and save time. Even bigger gains can be made by employing more automatic computing tools such as GNU 'make', 'git' version control, 'markdown' for reporting and R packages like 'dryworkflow' and 'ProjectTemplate'. These tools help to implement a don't repeat yourself (DRY) approach. They also aid reproducible research which is the idea that journal articles are published along with the data and software code so that others may verify the findings and build upon them.
In this short talk, Peter will outline the growing interest in so called reproducible research and the ideas he presented in a half day tutorial and talk at the R Users Conference, Aalborg, Denmark in July 2015. The conference attracted over 650 participants and in addition to presenting many cutting edge statistical applications, also highlighted Microsoft's recent moves in incorporating R into SQL Server 2016 and the Microsoft Azure cloud computing platform. Further details may be found at www.petebaker.id.au.