kbagstad
commented
12 years ago Also, just got model -d results, and two strange things happen - first, no data from the riverfront layer, though it checks out OK when doing a model -o, when doing a model -d just of riverfront, and when doing a rank -d of aestheticService:RiverfrontPresence.
Both model -d and -o report the highest value for the source value is around 37.48, well below what actually should be the highest values.
So I'm not totally sure the riverfront layer's being treated properly/consistently. Also, note that the "OpenSpacePotentialValue" is a NoisyMax node. Not sure if that could cause strange behavior, since in some cases we're setting BOTH park and forest as present (i.e., is this OK for a NoisyMax?)
Just some thoughts.
@fvilla
OK, here's another example of "reversed values" (OK, I know they're not reversed, but that's how they're appearing) like what we had on the San Pedro recreation models back in the day. Note that I've updated and pushed new colormaps.properties files that should contain the proper ordering for any ordered concepts with ambiguous names, and also for presence/absence so they display consistently.
The clearest place to see the results coming out wrong is where Olympic National Park and National Forest meet the Quinault Indian Reservation (see http://nationalatlas.gov/printable/images/pdf/fedlands/WA.pdf for reference). If you run model -o proxwa.nc core.models.aesthetic-proximity-puget/source core.contexts.puget/puget2048, you'll see common values of 30.214 on private lands outside the NP, NF, and Reservation, common values of 27.520 on the reservation, and 21.314 in the National Park and Forest.
To set the BN ProximitySourcePuget.xdsl correctly for input data (which all look good from the .nc file), you'll use values of: VeryLargeArea, NoSurfaceWater, LowCrimePotential, ForestPresent, and Lakefront, Farmland, Riverfront, Beach, EmergentWetland, and WoodyWetland all absent. Cemetery and GolfCourse use priors. That gives you your probability distribution for private lands outside the NP, NF, and Reservation - 31% high, 28% moderate, 30% low, 10% no - which should translate to a value of 38.5 (yet values are 30.2). Next, set FormalProtection to "Protected" - this is the case of the Indian Reservation, and gives values of 47% high, 29% moderate, 20% low, 5% no, which should translate to 49.25 (yet values are 27.5). Finally, keep FormalProtection as "Protected" and switch Park to "ParkPresent". This should give values of 64% high, 23% moderate, 11% low, 3% no, which should translate to 58.35 (yet values are 21.3). So you can see how values are "reversed".
Thanks for looking into this.