Biostatistics
The field of biostatistics involves development and implementation of statistical methods to answer scientific research questions based on quantitative data. To investigate and document effects of new medicines, new treatments and changes in lifestyle behaviors, large amounts of information are routinely gathered in a modern health care system. This information is often based on clinical data, data from the laboratory or data from national health care registers.
Biostatistics is a necessary tool to make valid interpretations and conclusions based on such data. The methods we use provide an overview of the data and explore relevant associations between variables in a succinct and accurate manner, and our results are always presented together with an estimate of how certain we are of them.
As researchers in the field of biostatistics, we develop new methods, evaluate their properties and apply them in new situations. We participate in research projects across all departments at Health as well as we work together with other research institutions. This broad perspective generates new research ideas and ensures that we are up-to-date with state of the art methodology.
We share our knowledge with the coming generations of researchers at PhD courses at the faculty.
Methods
The methods we currently work with include:
- Causal inference
We develop methods to estimate causal relationships between variables based on observational data. We evaluate asymptotic properties of estimators of various causal quantities. - Correlated ordinal data
We develop statistical models for correlated ordinal data, typically obtained when an underlying continuous outcome cannot be measured. Focus is on statistical properties of the distribution of the underlying continuous variables based on the multivariate ordinal data. - Multiple event data
We develop methods for analysing prioritised event data based on censored time-to-event data such as the win ratio. This includes estimation and regression models for the win ratio. - Pseudo-observations
This method estimates adjusted associations between e.g., exposure and survival, and it is suited for situations where participants have varying follow-up time. We develop methods for scenarios such as delayed-entry and survey sampling.
