Research projects at the Department of Statistics
Overview about the research projects at the Department of Statistics.
Mining for Meaning: The Dynamics of Public Discourse on Migration
This research environment seeks to understand how Sweden is adapting to modern migration and the integration of the newly-arrived. Bringing together relevant scholars with diverse methodological and substantive expertise, we combine cutting-edge research on migration and integration with the development of machine-learning methods for the analysis of text in the social sciences.
Welfare State Analytics
Welfare State Analytics. Text Mining and Modeling Swedish Politics, Media and Culture, 1945-1989 (Westac) is a digital humanities research project with five co-operatings partners: Umeå University, Uppsala University, Aalto University (Finland) and the National Library of Sweden. The project will digitise literature, curate already digitised collections, and perform research via probabilistic methods and text mining models.
Development and clinical validation of sepsis prediction algorithm in an ICU
The aim of the project is to first improve and then clinically validate an algorithm for predicting sepsis. The project stakeholders will work together to refine the algorithm, develop a user interface and test the algorithm in a clinical setting in an intensive care unit (ICU). The algorithm is being developed using so called machine learning in a collaboration between AlgoDx AB and Uppsala University.
Inference of Causal Effects from Complex Longitudinal Data Based on the New G-Formula
This project will develop a new efficient approach for sequential causal inference based on the new G-formula in the framework of single-point causal inference. Furthermore, the research team will apply the method to stomach cancer patients’ clinical data and find optimal individual treatments and treatment regimes for stomach cancer.
New Statistical Methods for Latent Variable Models
This research project aims to develop new efficient statistical methods to better model linear and nonlinear relationships and accurately predict latent variables within LVMs. The hierarchical-likelihood, a generalization of Fisher's likelihood, will serve as a fundamental tool in this project. Within the h-likelihood framework, we propose estimation methods and inference procedures for various LVMs. To our best knowledge, h-likelihood has never been applied to the models in this project. H-likelihood will be thoroughly studied under different conditions. Further, it will be applied to the nonlinear model and ordinal data for the first time.
Significance of control variables with measurement errors in register data when estimating causal effects
The project aims to study systematic errors that arise due to measurement errors of control variables in estimates of causal effects, which is especially important when control variables are retrieved from register data. More specifically, the project aims to study a specific type of measurement error that is common, namely one-sided error classification. This is a measurement error that occurs, for example, due to under-reporting in registers. This means that individuals who have a registered property actually have the property, but there are also individuals in the register who have the property without the property being registered.