| title: | Analysis of medical data with incomplete information |
|---|---|
| reg no: | ETF5203 |
| project type: | Estonian Science Foundation research grant |
| subject: |
1.2. Applied Mathematics 3. Medical Sciences |
| status: | completed |
| institution: | University of Tartu |
| head of project: | Krista Fischer |
| duration: | 01.01.2002 - 31.12.2005 |
| description: | The aim of the project is to extend existing methods and develop new methodology for the analysis of medical data, when the data does not contain complete information on quantities of interest. There are at least two general cases that one can consider as the problem of incomplete information. First, quite often in the practice of statistical data analysis one encounters the fact that not all intended data is observed: some subject records are incomplete ore missing. The problem is especially frequent in longitudinal studies. By ignoring missing data, the resulting estimates are biased for the population parameters of interest. The aim of the project is to investigate the existing methods to correct these biases in concrete epidemiological studies and to develop new methodology. The issue of incomplete infiormation arises also by the problematics of causal inference. In medical context, one is often interested in the causal effect of actually received treatment, but is not able to distinguish it from the selection effects caused by unknown confounders. The methodology of structural mean models leads to unbiased estimates in the presence of selectivity and is developed for placebo-controlled trials. The plan is to extend that methodology for more general cases: trials comparing two active treatments, trials with repeated measurements and observational epidemiological studies. In the last case the plan is to use also the results from econometrics, mainly instrumental variables estimators. Besides estimation methodology, it is necessary to develop tools for model diagnostics in all contexts mentioned above. This is the least developed aspect upto now in the analysis of incomplete data and causal inference. The aim is to develop appropriate test-statistics and investigate their theoretical properties together with some graphical tools. As the result, it is expected to publish several scientific articles in international journals and obtain useful information from epidemiological and medical data analysis that is important from medical and public health aspects. |
| project group | ||||
|---|---|---|---|---|
| no | name | institution | position | |
| 1. | Krista Fischer | Faculty of Medicine | associate professor | |
| 2. | Ene Käärik | Institute of Math. Statistics | lecturer | |
| 3. | Katrin Lang | Institute of Public Health | senior assistant | |
| 4. | Märt Möls | Institute of Math. Statistics | researcher | |
| 5. | Heti Pisarev | Department of Paediatrics | statistician | |