A Bootstrapping Based Simulation Approach for a Problem from Neonatology

Statistical inferences based on small dimension samples represent a big problem and a made to measure challenge. In the biomedical domain there are many situations when costs or ethical reasons enforce that only a few data are collected. Nevertheless, inferences must be made. In this paper, an alternative to the classical statistical approach is discussed. The article is focused on a simulation strategy through bootstrap re-sampling. Linked to this, an important phenomenon appears during bootstrapping, namely bootstrap aggregation. This is also revealed through simulation. The model is tested on real small samples of medical data concerning the oxidative stress at newborns, in term and premature, comparatively. All calculations implied by the model are implemented through Matlab scripts.