Introducción y aplicación metodológica de las puntuaciones de propensión: Revisión bibliográfica y simulación de datos
The purpose of this paper is to introduce propensity scores (PS) as a key statistical methodology to reduce confounding bias in observational studies and thus facilitate causal inference. In contexts where randomisation is not possible for ethical or practical reasons, such as in clinical or epidemi...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2025 |
| País: | España |
| Institución: | Universitat Oberta de Catalunya (UOC) |
| Repositorio: | O2, repositorio institucional de la UOC |
| OAI Identifier: | oai:openaccess.uoc.edu:10609/151870 |
| Acceso en línea: | https://hdl.handle.net/10609/151870 |
| Access Level: | acceso abierto |
| Palabra clave: | propensity scores causal inference observational bias puntuaciones de propensión inferencia causal sesgo observacional Biometry -- TFM Biometria -- TFM |
| Sumario: | The purpose of this paper is to introduce propensity scores (PS) as a key statistical methodology to reduce confounding bias in observational studies and thus facilitate causal inference. In contexts where randomisation is not possible for ethical or practical reasons, such as in clinical or epidemiological trials, this approach is particularly relevant. An exhaustive review of the literature is carried out, explaining concepts, application methods (matching or weighting, among others) and libraries available in different programming languages. A case study is also implemented using simulated data in R to evaluate the effectiveness of PS against traditional methods such as regression. The results show that PS allows balancing covaraites between treated and untreated groups, significantly reducing bias and improving the quality of causal estimates, both ATT and ATE. Appropiate use of PS was found to minimise baseline imbalance and improve the internal validity of the studies. In conclusion, PS represents a robust and versatile methodology to address bias problems in observational studies. It also highlights the importance of validating and optimising the adjusted PS model to ensure accurate and reproducible results, suggesting future lines of research and including applications in real clinical trials. |
|---|