Flexible modelling techniques and use of historical controls in animal studies

Abstract

This thesis has touched upon two broad sections, the main part being flexible modelling techniques, with an in-depth focus on penalized spline methodology and the use of historical data in animal studies. We have demonstrated the versatility of the penalized spline based methodology. In particular, the use of the same basic model in different study designs and different intricacies is quite appealing. This follows the application of the methodology with continuous data to the parallel design case in Chapter 4, and also for the cross-over design in Chapter 5 and Chapter 6. The intermingling of the penalized spline methodology with surrogate marker validation techniques in Chapter 6 demonstrates its widespread applicability. Use of similar methodology with non-normal data was illustrated in Chapter 7 and Chapter 9. The different models we proposed, coming from manipulating how penalized splines are constructed are applicable in many different scenarios. While the main inferential tool for data of a longitudinal nature was simultaneous confidence bands, accounting for various sources of variability, it will be interesting to compare the results obtained with other methods, for example, the approach of Behseta and Kass (2005)....