Statističke metode u planiranju farmaceutskih ispitivanja

Radnić, Marija (2016) Statističke metode u planiranju farmaceutskih ispitivanja. Diploma thesis, Faculty of Science > Department of Mathematics.

[img]
Preview
PDF
Language: Croatian

Download (1MB) | Preview
[img] Archive
Language: Croatian

Download (2MB)

Abstract

The successful research and development of new drugs is critical for the future of the pharmaceutical industry and for human health. One particular aspect of drug development that pharmaceutical companies are required to consider is the ability to identify subgroups of patients that are likely to derive additional benefit from treatment and to help quantify the burden of those patients on the healthcare system. One of the methods that can be used is PRIM method (Patient Rule Induction Method) as shown on the example in section 4. PRIM is intended as an addition to the data analyst’s tool kit, to be used when the goal is either explicit or implicit optimization (section 2.2). Its distinguishing characteristics include patient peeling/pasting coupled with multiple trajectories (section 2.5) to enhance power and stability, and intimate user involvement in the model selection process. It tends to produce parsimonious interpretable descriptions of the structure it uncovers, and on the problems considered here exhibits performance superior to comparable procedures such as CART (Classification and Regression Tree Method) that are intended for function approximation. The extent to which this performance gain generalizes other situations remains to be established. As with all learning procedures, relative performance will likely be problem dependent. The final chapter gives an example of application PRIM method for analyzing the risk of IHD.

Item Type: Thesis (Diploma thesis)
Supervisor: Slijepčević, Siniša
Date: 2016
Number of Pages: 48
Subjects: NATURAL SCIENCES > Mathematics
Divisions: Faculty of Science > Department of Mathematics
Depositing User: Iva Prah
Date Deposited: 11 Apr 2016 11:27
Last Modified: 11 Apr 2016 11:27
URI: http://digre.pmf.unizg.hr/id/eprint/4687

Actions (login required)

View Item View Item