Open Jamesrrobbins opened 1 year ago
Hi James, thanks for the feedback and for the useful suggestion.
This app is currently undergoing an updating process (e.g. inclusion of multiple WF scenarios, faster runs, etc), where the option of selecting the starting seed is available.
The Beta version is here - https://github.com/dmpstats/sCRM. Feel free to take it for a spin!
Hi Bruno, Thanks for the information, and for sharing the beta! I've had a bit of a look around the beta and it sounds like it will be incredibly useful - a very exciting upgrade to an already slick app. Being able to run multiple scenarios at once will be of huge help to many - I'm quite often running 30+ turbine design possibilities! One other addition that I've made to a local version of the app is to make it write out the distribution of collision estimates that summary statistics are calculated from - we'll be using this for some cumulative collision risk work across multiple wind farms. Potentially not that wide-spread a need for it, but perhaps worth considering too. Looking forward to seeing this new version rolled out fully in the future. Thanks, James
On Fri, Nov 25, 2022 at 1:29 PM Bruno Caneco @.***> wrote:
Hi James, thanks for the feedback and for the useful suggestion.
This app is currently undergoing an updating process (e.g. inclusion of multiple WF scenarios, faster runs, etc), where the option of selecting the starting seed is available.
The Beta version is here - https://github.com/dmpstats/sCRM. Feel free to take it for a spin!
— Reply to this email directly, view it on GitHub https://github.com/dmpstats/stochCRM/issues/29#issuecomment-1327481584, or unsubscribe https://github.com/notifications/unsubscribe-auth/AI54LQEMIU3DD2QLFQINCY3WKC5M3ANCNFSM6AAAAAASLIXK3A . You are receiving this because you authored the thread.Message ID: @.***>
Cheers James. The upgraded version also outputs all the draws of simulated collisions, so that should help.
For those large numbers of turbine scenarios I'd probably stick to the recently released stochLAB package - a bit more lead-up coding time required for setting up multiple scenario runs, but in the long term more efficient and reproducible when compared to typing inputs in the app, specially if re-runs are needed (as they almost always are!)
Great work on this app - I use it regularly. One thing that would be useful is the ability to reproduce results by adding functionality to select a starting seed.
I have coded this into the local shiny app for personal use, but think it would be useful to be rolled out for others for greater reproducibility and transparency.
All that I did was:
Added the following to L358/359 in ui.r, below selection for number of iterations:
hr(), numericInput("seed", "Starting seed", 52),
within server.r: observeEvent(input$actButtonInput_simulPars_GO, { set.seed(input$seed) ...
elsewhere in server.r (potentially not needed): reactive({ set.seed(input$seed)
})
I think this addition would help a lot of us in consultancy who end up running many iterations of these models over the development lifecycle of projects.