We show that in most cases, when the H0 is true, the BF fails to reach/exceed BF > 6. Following previous recommendations, we suggest the use of flexible thresholds. As such, even when researchers are aiming for a BF > 10 for H1, a BF > 3 should be considered adequate for H0. 5/5
We are not the first to mention this disanalogy, however we wanted to provide evidence for it. We run a simulation and calculated 200 million BF, simulating either a true H1 or a true H0 with various n, sd, and d. 4/n
For example, with the commonly used Bayesian t-test, there is disanalogous relationship between calculating a BF when H1 is true vs when H0 is true, when relying on the Savage-Dickey approach. 3/n
Bayes factors are continuous and updatable. However, some psychology journals define required thresholds (e.g., BF > 3, BF > 6, BF > 10). The same thresholds need to be surpassed in favour of either H1 or H0. We argue that this approach is problematic in some cases. 2/n
Thanks Matt! Glad you find our work neat!
Thanks Anna. This is brilliant advice, and probably the way to go!
I’m not really going anywhere with this, I just find it amazing, that one experience can be both amazing and suck at the same time. The only thing I know for sure, is that if I was given the opportunity to start over, I’d make exactly the same choices. End.
with the people who are truly close to me. Further, as a post-doc, I don’t really fit in. Students see me as more senior, and faculty see me as very junior. So, socially, it gets tough to fit into a group.
However, I had to leave everything behind and move to the other side of the world, for a temporary contract. As a first-gen academic, with none of my close friends having gone through the world of academia, I find it hard to explain what I’m going through (both the good and bad)
I have brilliant supervision with lots of freedom, decent compensation, and access to the equipment I need for my research. I get to try new things, exchange skills & expertise, do amazing research, and at the same time, not worry about rent and exams.