You should only use wide priors if you have no information about your data. It should usually be possible to come up with more informative prior distributions by reading other studies in your area or talking to experts.
Yes and no. With quality data, the prior is overwhelmed by the data, so the prior shouldn't matter too much as long as it's in the right ballpark. And translating abstract information from other studies and experts into a prior distribution is surprisingly challenging.
One of the bast classes ever! Congrats.
Just learned more in this brilliant lecture than the whole semester!! GRACIAS!!! ❤️🇲🇽❤️
brilliant man. Period.
Wow
Running bin_unif
I wish our econometrics professors were teaching like you. Outstanding!!
This was incredibly helpful, thanks for sharing this video!
Really helpful thank you very much!!
oh Waaw! I paid around a 10 dollars to learn this in my university and this guy does it better than anyone is had seen!
wow!! could be the best lecture i ever heard! thank you!!
Can you please provide the link to download the dataset used in the video?
Unfortunately, the data is confidential. So while I can share the analysis, I'm not allowed to share the raw data. Sorry about that.
Can you please clarify the fact the no multiple testing correction is needed for Bayesian approach - thank you
Great lecture, thanks!
hello, how can i define a prior from a previous experiment?
I know the beauty of Bayesian now!
Fantastic lecture. Thank you
wow this resolves my confusion between the 2
Great lecture!
You should only use wide priors if you have no information about your data. It should usually be possible to come up with more informative prior distributions by reading other studies in your area or talking to experts.
Yes and no. With quality data, the prior is overwhelmed by the data, so the prior shouldn't matter too much as long as it's in the right ballpark. And translating abstract information from other studies and experts into a prior distribution is surprisingly challenging.