NEWS A new thesis by Guilherme Wang de Faria Barros, Umeå School of Business, Economics and Statistics, focuses on challenges in time-to-event studies based on observational data, particularly in the context of stroke research using the Swedish Stroke Register (Riksstroke).
Doktorand vid Handelshögskolan vid Umeå universitet
ImageHans KarlssonThe primary interest lies in understanding the time to recurrent stroke or death after an initial stroke. The main parameter of interest is the hazard ratio, comparing risks between treated and untreated groups, often estimated using the Cox regression model.
Observational data, unlike experimental data, introduces challenges due to the presence of confounders—variables that can distort the relationship between treatment and outcome.
“The use of national registers facilitates access to rich observational data, and Sweden is a country known for its many high quality national registers relating to different topics regarding its population”, says Guilherme Wang de Faria Barros.
The four papers in the thesis address specific challenges in time-to-event studies. Paper I tackles the issue of competing risks in survival analysis, comparing individuals with and without modifiable risk factors using data from the Swedish Stroke Register (Riksstroke).
Papers II and IV delve into methodological aspects. Paper II explores non-parametric methods for balancing datasets, offering alternatives to traditional parametric methods. It includes a case study on anticoagulant prescription after a stroke. Paper IV addresses covariate selection in high-dimensional data, proposing methods to navigate situations where the number of covariates is comparable to the number of individuals in the study.
Paper III investigates the impact of censoring on marginal hazard ratio estimation. It explores solutions considering varying effect sizes and censoring rates, presenting a procedure to mitigate the issue.
“Although this thesis focuses on stroke, the topics covered in each paper address practical challenges in any observational time-to-event study, making the research relevant for improving methodologies and contributing to the field of observational data analysis”, says Guilherme Wang de Faria Barros.
The Department of Statistics, Umeå School of Business, Economics and Statistics, at Umeå University has numerous researchers focused on the study of causality and analysis of data from registers, and is a leading university in the topic.
About the public defence of the thesis
Guilherme Wang de Faria Barros, Umeå School of Business, Economics and Statistics, is publicly defending his doctoral thesis with the title: Estimation of hazard ratios from observational data with applications related to stroke Faculty Opponent: Daniel Nevo, Associate Professor, Department of Statistics and Operations Research, Tel Aviv University
