Quadratic inference function with ridge estimator for myopic regret-regression: The short-term strategy in optimal dynamic treatment regimes / Nur Raihan Abdul Jalil
A dynamic treatment regime (DTR) is a multi-stage decision rule based on treatment history. The focus in this research is on the improvement of estimation in optimal dynamic treatment regime (ODTR). This research is motivated by the regret-regression method where it is a combination of the regret...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/14734/1/Nur_Raihan.pdf http://studentsrepo.um.edu.my/14734/2/Nur_Raihan.pdf http://studentsrepo.um.edu.my/14734/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | A dynamic treatment regime (DTR) is a multi-stage decision rule based on treatment
history. The focus in this research is on the improvement of estimation in optimal dynamic
treatment regime (ODTR). This research is motivated by the regret-regression method
where it is a combination of the regret function with regression modeling. A short-term
strategy called the myopic regret-regression (MRr) is an alternative to regret-regression
where it estimates the mean response at each time-point. This strategy has the same
performance as regret-regression but MRr calculation is faster in estimation and more
practical in application. However, it has a limitation on correlated data. The quadratic
inference functions in myopic regret-regression (QIF-MRr) has overcome the limitation
of MRr by combining the myopic regret-regression with quadratic inference functions.
It is more robust and efficient regardless of any type of working correlation structure.
However, singularity problem happen when estimating the parameters using QIF-MRr and
more complex in computations. Hence, the ridge quadratic inference functions for myopic
regret-regression (rQIF-MRr) is proposed where a ridge estimator is used to overcome the
computational problem and shorten the calculation time. Comparison between methods
was performed to check the efficiency and consistency in estimation using simulation with
different sample sizes.
|
|---|