Open karimn opened 4 years ago
@wwiecek The run on the cluster using the narrower prior for imputed cases is still producing some high Rhat/low ESS for some countries like Canada and India and some countries like Poland and Peru look like they might doing worse (this could be just random). Anyway, I think I'll try to see if centering improves things for at least some of the countries.
Good idea, centered can be better in some situations.
Would be good to see what happens in these new runs vs what we had previously, esp for Poland or Peru. As I said yesterday, putting a narrower prior everywhere may lead to problems in places which genuinely had a lot of initial seeding. Are the results from that new run pushed somewhere?
Pushed pl*.RData and pe.RData to data/mobility/results on main-karimn*
On Thu, Jul 2, 2020 at 4:59 AM Witold Wiecek notifications@github.com wrote:
Good idea, centered can be better in some situations.
Would be good to see what happens in these new runs vs what we had previously, esp for Poland or Peru. As I said yesterday, putting a narrower prior everywhere may lead to problems in places which genuinely had a lot of initial seeding. Are the results from that new run pushed somewhere?
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/karimn/covid-19-transmission/issues/17#issuecomment-652882590, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABB556MRLDTH5G227JJOMLRZRD55ANCNFSM4ON4GZDA .
Thanks -- I fetched but then today got really busy and I did not get a chance to have a look
Update: debugging through new code allowing us to turn on/off centering of R0 and beta parameters. Taking a bit of time as the centered parameters are always more complicated to implement, I feel. Work underway in branch karimn/center-param.
On Thu, Jul 2, 2020 at 4:37 PM Witold Wiecek notifications@github.com wrote:
Thanks -- I fetched but then today got really busy and I did not get a chance to have a look
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/karimn/covid-19-transmission/issues/17#issuecomment-653209704, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABB55Z6BOM6QOZVUQNXASDRZTVXVANCNFSM4ON4GZDA .
All hierarchical parameters are currently uncentered. I was reading this discussion https://discourse.mc-stan.org/t/low-ess-and-high-rhat-for-random-intercept-slope-simulation-rstan-and-rstanarm/9985/6 and thought I should try adding the option to have these param centered and see what happens to our runs.