It is Xmas again! Wow, how quickly time flies. The situation here in the Netherlands is not ideal as the number of infected is on the rise again and the government has declared a full lockdown until January 19th. Starting from next month I will start teaching in the department but everything will still be online due to the uncertainty of the pandemic. It is unfortunate but there is nothing we can do about it and let us just hope the situation will improve in the next few weeks/months.
Aside from sad stuff, I have some good news about my research. My paper about longitudinal models in trial-based CEA, whose pre-print is available on my ArXiv account, has been officially accepted for publication in Value in Health, after a very long time (more than 2 years of peer-review process I believe!). I am really happy as this is a paper for which I spent a lot of time and effort, so it is nice to see some ouput out of it. I am also glad to announce that a new paper I am co-authoring with Baptiste (LSHTM) and Catrin (Bangor University) which is a tutorial on the use of mixed effects models for trial-based CEAs. The main idea is to make available and spread awareness in the HE community about the possibility to use these models to derive CEA results, rather than relying on the standard linear regression models. The methods are nothing special but we belive health economists are not well aware of the fact that sometimes they can be implemented in a much easier way compared to standard appraoches. We tried to summarise the main advantages and disadvantages of the methods while also providing software code (in STATA and R) to show how they can be fitted. The paper has been accepted as a contributed presentation at the next HESG meeting. Unfortunately, I will not be able to attend the meeting as I will be quite busy with the teaching that week but Catrin has kindly agreed to present in behalf of all authors. I look forward to receive the feedback from the people attending the conference!
The past month I have been quite busy with some teaching and consultancy work here at UM, mostly for medical students about basic statistical concepts and techniques but nonetheless very interesting for me in order to get acquainted with the new job and duties. A big thank is due to my new colleagues who have been incredibly nice to me and have helped me a lot to get into the system. I was actually planning to post my news earlier this month but with incredible disappointment I found out that the most recent updated of Hugo (the system used by blogdown for building this site) completely messed up my previous version of the website which was broken. I had to rebuild the website again copying and pasting all my previous material. Nothing crazy, but certaintly very annoying to do. I just hope the next update will not force me to do it again! So, if you see something different compared to before, you know why (especially in the sections for the tutorials on using BUGS/JAGS/STAN). Related to this, I must acknowledge Mr. Kim (firstname.lastname@example.org) who recently contacted me with regard to an incorrect specification for one of the STAN models in the tutorial on generalised linear mixed models. I would like to thank him again very much for noticing this mistake which I fixed in the new version of the website now online (apologies for the long time it took for me to make the changes!). I am happy to see that my code can be of help to anyone who might be interested in doing some nice modelling.
Bayesians: a request from @TheStatsGeek to explain what a credible interval actually means.— Tim Morris (@tmorris_mrc) November 22, 2020
Personally I’d like to see something beyond a pithy ‘it’s the prob that’ or ‘tells you how to bet’, without appealing to freq. calibration.https://t.co/g8ZtH4VaT3
where he asks about the interpretation of a Bayesian credible interval while also saying that quick and simple answers will not be considered reliable. Well, I have much to say on this but I feel like twitter is not the best location to try and argument a proper discussion. I hope I will be able to find some time to post on my website a more constructive answer. The topic is not very quick to grasp, especially if someone is used to think in frequentist terms and theory which, by definition, are not very useful to think at statstics from a Bayesian point of view. For the moment, as a quick answer, I would just say that in my opinion statistics is not a uniquely defined discipline but there are different ways it can be approached, and the rules and theory of one approach do not necessarily apply to others!
Next time, perhaps, I will follow this with a proper argument, stay tuned!