Abstract
New computational methods have recently been developed that allow stable fitting of constrained GLMs with bounded non-canonical link functions, such as the log link binomial model. By employing B-splines, we can extend these approaches to allow for semi-parametric adjustment of rate differences, risk differences and relative risks. These methods provide alternatives to standard fitting methods, resulting in greater stability for accommodating the required parameter bounds. They also provide a straightforward way to accommodate additional restrictions such as monotonic regression functions. We demonstrate an application to data from a clinical trial of oxygen supplementation in premature infants.
Original language | English |
---|---|
Title of host publication | Proceedings of the 29th International Workshop on Statistical Modelling |
Editors | Thomas Kneib, Fabian Sobotka, Jan Fahrenholz, Henriette Irmer |
Place of Publication | Göttingen, Germany |
Publisher | Statistical Modelling Society |
Pages | 105-110 |
Number of pages | 6 |
Volume | 1 |
Publication status | Published - 2014 |
Event | International Workshop on Statistical Modelling (29th : 2014) - Göttingen, Germany Duration: 14 Jul 2014 → 18 Jul 2014 Conference number: 29th |
Conference
Conference | International Workshop on Statistical Modelling (29th : 2014) |
---|---|
Abbreviated title | 29th IWSM 2014 |
Country/Territory | Germany |
City | Göttingen |
Period | 14/07/14 → 18/07/14 |
Keywords
- Generalized additive model
- Semi-parametric model
- Rate difference
- Risk difference
- Relative risk