Some updates ...
Hello guys, I hope everything is well with you and that you had some nice Easter break (at least I did). Personally, I really needed this break as I was quite tired and without energy due to the heavy workload in the past months. Now I feel refreshed and ready to start anew (let’s see how long this can last). Anyway, today I wanted to give some updates about what is up for me in the upcoming months as it has been a while that I have not made a post like this.
First of all, I am happy to announce that my working paper has been accepted fot discussion at lolaHESG, which is the corresponding Dutch conference to the popular Health Economics Study Group conference in the UK, dedicated to researchers involved in health economics’ topics. I already attened last year’s edition, held in Maastricht, which was a blast. Amazing people and colleagues to talk to, nice food and accommodation, and most importantly lots of different presentations to attend and opening up of new connections and work relationships. Similarly, to last year, also for this year I am going to present some on-going work focussed on the modelling of health economics trial-based data which takes inspriation from one of my most recently published papers about the use of longitudinal linear mixed models for the analysis of health economics trial data. The paper was a simple introduction to practitioners to the use and specification of mixed models for the analysis of trial-based health economics data, with a practical illustration on how to specify the models in software such as STATA or R, full code available on my GitHub. In the work I am going to present at lolaHESG, I hope can start from this introductory and applied paper and perform some simulations to assess the performance of the proposed methods across different missingness and data structure scenarios. It takes a bit of time but I think it will definitely be worth it, at least so that I can get some feedback form the other attendees on how to extend the work further (perhaps for a future paper). I am finalising the work in these days but I need to hurry up as the deadline for the submission of the full paper is the end of this week, I hope I can make it! I also have some nice ideas to further extend the work but I need to have some feedback on what I have done so far first.
Second, and a bit unrelated to what I said, I wanted to breifly comment a recent tweet from Gelman
4 different meanings of p-value (and how my thinking has changed) https://t.co/xY7vFX3IYf— Andrew Gelman et al. (@StatModeling) April 14, 2023
which I find it extremely interesting from a purely statistical perspective (sorry health economist audience!). The tweet links back to his blog on a post about a discussion on the concept and definition of the term p-value under a classical statistical framework which, unlike what we are told in basic statsitics course, is not so clear-cut across all scenarios. Indeed, he refers to four different alternative definitions of p-values, which vary depending on the specific quantity the researcher refers to with the concept of p-value. In some scenarios, and typically these are the cases encountered in basic analysis problems, these definitions coincide and we have our usual description of a p-value. However, when we go outside such simple scenarios and focus more on what we mean by the term p-value, e.g. based on an test statistic or family of hypothesis tests. This is very interesting to me as the already quite confusing concept of a p-value and its relevance in the context of hypothesis testing framework is further subjected to easy misunderstandings. The whole discussion is about how can we define in an unique way the concept of p-value and the answer, as it usually is in these cases, is that there is no unique definition as it depends on what we mean by the quantity that we define with p-value, e.g. probability statement or a general summary of the data without any probabilistic meaning attached to it. As also Gelman points out in his post, also my thinking about p-values has changed over time. From necessary evils to be used to achieve some statistical conclusions about the likely existence of an effect to until confusing terms that people tend to misinterpret and based on which no critical decision should be based blindely. Regardless of this rumble of thoughts about p-values, I too think that it is not a problem that p-vaules are used in their common interpretation even though a general an unique defintion for such term may not be possible. What is important I feel is that people are aware of the possibly confusing term that a p-value is and that they need to not blindly trust and interpret such values as if they were true statements, as what is meant by such statement may not be fully clear from the start.
Let’s think carefull when we do data analysis!