Closed elbfox closed 4 years ago
This issue is a little more nuanced than I thought. I've been doing multiple rounds of testing with equal prefs because it helped catch a lot of little idiosyncrasies in the normalization/results output. With equal preferences, it looks like the ind/multi-dam results diverge dramatically (see edited issue text above). With fish preferences ( FishPrefs_forLiveSite.xlsx ...don't forget to save and upload as CSV) the results are the same for ind/multi-dam (Yellow = Remove Dam.
With hydro preferences (HydroPrefs_forLiveSite.xlsx) the results diverge again: This one is harder to read, so this is top decision alternative version: As you can see, the map doesn't show just "improve hydro" and "keep and maintain" options, but rather shows "Fish AND Hydro" and "keep and maintain".
Hypothetical Indigenous prefs file (IndigPrefs_forLiveSite.xlsx ...developed as an example for the NOAA workshop) where the multi-dam results are consistent with what we'd expect based on the ind. dam results.
And, perhaps most interesting yet, the lakeside homeowner prefs file (HomeownerPrefs_forLiveSite.xlsx.xlsx) Where the results don't quite match up (e.g., Dolby and North Twin would be the ones with fish passage improvements based on the individual dam results above)
So, is it possible that this is just because the social criteria values are different between ind. dam and multi-dam datasets right now? Or is there something else going on? The criterion that seems to be driving the differences is Sea-Run Fish Habitat Area, which we had anticipated due to the network-dependent nature of the data, but it just seems like when there are differences, the differences are considerable, or enough to make you wonder what is going on. I think we will need to wait for the new MOGA results (@elbfox to email Sam the new DamsData_workshop.csv) and the most recent PR to be completed to test again @sythel), but it would be really helpful to have your perspective on this, Sam.
This may have something to do with it in addition to the Properties thing that Sam found this afternoon...just added this to server.R DecisionsFix <- Decisions DecisionsFix[,1] <- Decisions[,8] DecisionsFix[,2] <- Decisions[,7] DecisionsFix[,3] <- Decisions[,6] DecisionsFix[,4] <- Decisions[,1] DecisionsFix[,5] <- Decisions[,4] DecisionsFix[,6] <- Decisions[,3] DecisionsFix[,7] <- Decisions[,2] DecisionsFix[,8] <- Decisions[,5] Decisions <- DecisionsFix
PR #171 did not fix the issue. I am at a loss. the model output doesn't even match the right map anymore.
Excel Equal Prefs file (save as a CSV if you use to test on the website) EqualPrefs_forLiveSite.xlsx Individual dam output using Equal Preferences shows: Remove West Enfield Remove Medway Remove East Millinocket Improve Fish Passage Dolby Improve Fish Passage North Twin Remove Millinocket/Quakish Improve Fish Passage Millinocket Lake Remove Ripogenus The graph in the Combined Results page that shows these results: And the zoomed in graph to see the top decision alternatives for each dam:
Here is the Excel MCDA Check Calcs file so you can compare for yourself. MCDA_Check_CalcsGraphsEQUAL092519 (2).xlsx
Multi-Dam output using Equal Preferences shows MAP 994 (visual matches up with these decision alternatives, very different from the individual dam results calculated using the DamsData. file here (DamsData.xlsx) Note: this DamsData file is NOT THE SAME as what we will use in the workshop...this one is old and will diverge slightly from the version you used in the MOGA, @samGroy Fish Passage AND Hydro West Enfield Fish Passage AND HydroMedway Fish Passage AND HydroEast Millinocket Fish Passage AND Hydro Dolby Fish Passage AND Hydro North Twin Fish Passage AND HydroMillinocket/Quakish Fish Passage AND Hydro Millinocket Lake Fish Passage AND Hydro Ripogenus
One possible source of difference is in the social criteria values at this point (to be shared early on Monday, I've asked @sharonklein to take review first), because it has been a while since we updated the MOGA dataset, and we've had a few changes to those values, as well as the GHG Emissions criteria values. However, even with these data differences, I find it hard to believe that the outcome would be SO different, and also consistent for every dam. @samGroy I've set you as an assignee so you can see this GitHub issue and help us think about what to do. We need all the help we can get to solve this ASAP!