Hello everybody, it is time for some quick updates about myself and what I have been doing the past month. Well, essentially, I have been crazy busy doing lots of teaching on statistics-related subjects, which is the primary reason I was hired here in Maastricth. The students are mostly from undergraduate medical or health science programmes and therefore the level of statistics that I need to teach them is rather basic, although the crucial aspects of data analysis is very well-treated in the courses. So far I had a wide variety of students with different backgrounds and from my first impressions it seems that those who are younger, say in their first undergraduate year, have shown more interest and effort to learn compared with their older colleagues. However, I have been teaching only two courses here at UM and many more are coming up in the next few weeks/months. The statistical concepts covered are the usual frequentist ones, such as p-values, confidence intervals, regression analysis, etc… So, nothing incredible but since I was a student too some years ago I can related with some of the questions they have about these concepts, especially given that statistics is not exactly their main programme focus (although it is essential to learn it!).
Of course, since we are living in this moment, all teaching is done online and this makes things a little more difficult sometimes but in general I think I managed ok for answering the students' questions and providing them with feedback on their exercises and group works. It is quite annoying that this being my first time teaching these courses (already prepared by other colleagues) I have to spend a decent amount of time to prepare all the lectures and be ready to the many possible questions the students may have on each topic. This, however, was expected from the beginning and it is part of the job so I cannot complain too much!
Apart from my rumblings about teaching duties killing me, I have some other good news on my research activity, finally! I am happt to announce that my paper on longitudinal models for dealing with missing data under MAR in trial-based cost-effectiveness analysis is in press by the journal Value in Health and can be accessed at the following link. I am really happy about this publication as this took lots of time and effort and seeing this article being published is something that encourages me a bit as a reminder that all the efforts done has now repaid. Although joint longitudinal models are not exactly a novelty in statistics or even health economics, the way I present them here is kind of unique in that I do it from a Bayesian perspective (who would have guessed?) and I compare the performance of the method with respect to other more standard missing data appraoches such as complete case analysis, single and multiple imputation. The key message is that any method which discards some observed data is inevitably less efficient and more prone to bias compared with methods that do not ignore these data. Thus, given the characteristics of trial-based analysis, it is recommended that longitudinal models, which properly account for the time dependence between outcomes and use information from all collected observations, are used to minimise the impact of missingness assumptions on the conclusions of the study. These models can also be fitted using multiple imputation although, in my perspective, the main issue of that is that results must then be combined with bootstrapping in order to quantify the impact of uncertainty on the conclusions (while this is not necessary in a Baysian analysis!).
Hopefully, analysts will find the methods I propose in this article interesting and useful and they will agree that using simple but likely biased methods should not be the optimal choice when approaching to a trial-based analysis. Ok, I think that is all for the moment and it is time to go back to my teaching and (when I have time) researching!