// ~->[DNET-1]->~ // File created by someone at MonashUniv using Netica 2.06 on Jun 02, 2002 at 20:08:24. bnet syrotuck { autoupdate = TRUE; comment = "Syrotuck Bayesian Network for Lost Person Behavior\n\ -------------------------------------------------------\n\ This network is based on the data presented in William Syrotuck's booklet,\n\ \"Analysis of Lost Person Behavior,\" available from Barkleigh Productions, Inc.,\n\ through NASAR's bookstore (www.nasar.org). Originally published in 1976, I \n\ worked from the 2000 NASAR reprint.\n\n\ Syrotuck collected 242 cases of \"persons lost in wilderness areas.\" (Syrotuck\n\ counted his 13 \"Special category\" (\"despondent\" or \"mentally challenged\") \n\ cases separately, and reported 229 cases. His cases break down as follows:\n\ \|\t117\t\tWashington state\n\ \|\t 95\t\tNew York state\n\ \|\t 17\t\tIdaho, Oregon, California, Alaska, New Mexico, Wyoming, and Tenn.\n\ \|\t---\n\ \|\t229\t\t(so we don't know where the 13 special cases come from)\n\n\ Syrotuck found no useful difference in the median distances between New York\n\ and Washington. (Either the values were the same, or within 0.2 miles)\n\n\ He is presuming forested wilderness areas.\n\n\ I have documented my interpretations and assumptions in the descriptions for\n\ each node.\n\n\ -Charles Twardy\n\ April 2002\n\ "; whenchanged = 1023012489; visual V2 { defdispform = BELIEFBARS; nodelabeling = TITLE; NodeMaxNumEntries = 50; nodefont = font {shape= "Arial"; size= 10;}; linkfont = font {shape= "Arial"; size= 9;}; windowposn = (5, 7, 817, 612); CommentWindowPosn = (46, 46, 722, 458); resolution = 72; drawingbounds = (1218, 905); showpagebreaks = FALSE; usegrid = TRUE; gridspace = (6, 6); PrinterSetting A { margins = (1270, 1270, 1270, 1270); landscape = FALSE; magnify = 1; }; }; node Category { kind = NATURE; discrete = TRUE; states = (Retarded, Despondent, SmChild, Child, MiscAdult, Elderly, Hiker, Hunter); statetitles = ("Mentally Challenged", , , , , , , ); parents = (); probs = // Mentally Challen Despondent SmChild Child MiscAdult Elderly Hiker Hunter (0.0330579, 0.0206612, 0.0909091, 0.0991736, 0.0619835, 0.0991736, 0.1818182, 0.413223); whenchanged = 1020053521; visual V2 { center = (480, 78); height = 2; }; }; node Age { kind = NATURE; discrete = FALSE; states = (SmChild, Child, Teen, Adult, Elderly); levels = (1, 6, 12, 19, 65, 110); parents = (Category); probs = // SmChild Child Teen Adult Elderly // Category ((0.2, 0.2, 0.2, 0.2, 0.2), // Mentally Challen (0.05, 0.05, 0.4, 0.4, 0.1), // Despondent (0.95, 0.05, 0, 0, 0), // SmChild (0.05, 0.9, 0.05, 0, 0), // Child (0, 0, 0.18, 0.82, 0), // MiscAdult (0, 0, 0, 0.05, 0.95), // Elderly (0, 0, 0.18, 0.82, 0), // Hiker (0, 0, 0.18, 0.82, 0)); // Hunter ; comment = "Syrotuck gives ages for children and elderly, so this node\n\ lets the network infer category from age if it has to.\n\n\ I have made some edges a little fuzzy (so \"SmChild\" would\n\ have a 5% chance of having age \"Child\"). For Hiker, \n\ Hunter, and MiscAdult, I have split between \"Teen\" and\n\ \"Adult\" based on number of years: so teen covers 10 years \n\ out of the 55 years between 12 and 65, so gets 18%.\n\n\ "; whenchanged = 1020054123; visual V2 { center = (138, 66); height = 1; }; }; node Terrain { kind = NATURE; discrete = TRUE; states = (Flat, Hills); statetitles = (, "Hilly or Mountainous"); parents = (); probs = // Flat Hilly or Mountai (0.5, 0.5); comment = "Syrotuck splits terrain into \"Flat\" and \"Hill or Mountainous\".\n\ He gives conditional distributions for distance using both flat\n\ and hilly terrain, but does not say how hunters (for example)\n\ split between flat and hilly terrain.\n\n\ I have assumed a prior of 50-50 for terrain."; whenchanged = 1020053490; visual V2 { center = (144, 192); height = 3; }; }; node Dist_miles { kind = NATURE; discrete = FALSE; states = (One, Two, Three, Four, Five, Six, OverSix); levels = (0, 1, 2, 3, 4, 5, 6, 100); parents = (Category, Terrain); probs = // One Two Three Four Five Six OverSix // Category Terrain (((0.33, 0.42, 0.17, 0.02, 0.02, 0.02, 0.02), // Mentally Challen Flat (0.35, 0.32, 0.2, 0.04, 0.03, 0.03, 0.03)), // Mentally Challen Hilly or Mountai ((0.1428143, 0.1428143, 0.1428143, 0.1428143, 0.1429143, 0.1429143, 0.1429143), // Despondent Flat (0.1428143, 0.1428143, 0.1428143, 0.1428143, 0.1429143, 0.1429143, 0.1429143)), // Despondent Hilly or Mountai ((0.38, 0.46, 0.08, 0.02, 0.02, 0.02, 0.02), // SmChild Flat (0.63, 0.17, 0.09, 0.08, 0.01, 0.01, 0.01)), // SmChild Hilly or Mountai ((0.33, 0.42, 0.17, 0.02, 0.02, 0.02, 0.02), // Child Flat (0.35, 0.32, 0.2, 0.04, 0.03, 0.03, 0.03)), // Child Hilly or Mountai ((0.22, 0.41, 0.11, 0.065, 0.065, 0.065, 0.065), // MiscAdult Flat (0.37, 0.31, 0.16, 0.02, 0.02, 0.02, 0.1)), // MiscAdult Hilly or Mountai ((0.57, 0.28, 0.08, 0.0175, 0.0175, 0.0175, 0.0175), // Elderly Flat (0.52, 0.19, 0.19, 0.01, 0.01, 0.01, 0.07)), // Elderly Hilly or Mountai ((0.25, 0.25, 0.25, 0.0625, 0.0625, 0.0625, 0.0625), // Hiker Flat (0.29, 0.1, 0.27, 0.1, 0.04, 0.03, 0.17)), // Hiker Hilly or Mountai ((0.18, 0.47, 0.24, 0.0275, 0.0275, 0.0275, 0.0275), // Hunter Flat (0.31, 0.24, 0.2, 0.13, 0.03, 0.02, 0.07))); // Hunter Hilly or Mountai ; comment = "Straight-line distance in MILES from PLS to place found.\n\ Syrotuck gives his distance in miles. Easiest to just copy."; whenchanged = 1020053264; visual V2 { center = (678, 240); height = 7; }; }; node Vert { kind = NATURE; discrete = TRUE; states = (up, level, down); parents = (Terrain, Category); probs = // up level down // Terrain Category (((0.025, 0.95, 0.025), // Flat Mentally Challen (0.025, 0.95, 0.025), // Flat Despondent (0.025, 0.95, 0.025), // Flat SmChild (0.025, 0.95, 0.025), // Flat Child (0.025, 0.95, 0.025), // Flat MiscAdult (0.025, 0.95, 0.025), // Flat Elderly (0.025, 0.95, 0.025), // Flat Hiker (0.025, 0.95, 0.025)), // Flat Hunter ((0.33, 0.08, 0.59), // Hilly or Mountai Mentally Challen (0.5, 0.25, 0.25), // Hilly or Mountai Despondent (0.33, 0.11, 0.56), // Hilly or Mountai SmChild (0.33, 0.08, 0.59), // Hilly or Mountai Child (0.17, 0.17, 0.66), // Hilly or Mountai MiscAdult (0.1, 0.2, 0.7), // Hilly or Mountai Elderly (0.07, 0.04, 0.89), // Hilly or Mountai Hiker (0.06, 0.11, 0.83))); // Hilly or Mountai Hunter ; title = "Vertical Travel"; comment = "Did subjects go upward, stay at the same level, or go downard?"; whenchanged = 1020053087; visual V2 { center = (90, 348); height = 4; }; }; node FindLoc { kind = NATURE; discrete = TRUE; states = (Linear, Other); statetitles = ("Linear features", ); parents = (Category); probs = // Linear features Other // Category ((0.67, 0.33), // Mentally Challen (0.25, 0.75), // Despondent (0.57, 0.43), // SmChild (0.67, 0.33), // Child (0.3, 0.7), // MiscAdult (0.47, 0.53), // Elderly (0.73, 0.27), // Hiker (0.52, 0.48)); // Hunter ; title = "Find Location"; whenchanged = 1020052521; visual V2 { center = (276, 342); height = 5; }; }; node Weather { kind = NATURE; discrete = TRUE; states = (good, bad); parents = (); probs = // good bad (0.5, 0.5); comment = "Syrotuck estimates detectability in both good and\n\ bad weather, for each category. He does not say\n\ how many cases are in good versus bad weather,\n\ so I've assumed 50-50, and made it a root node."; whenchanged = 1020053095; visual V2 { center = (474, 420); height = 8; }; }; node Detectability { kind = NATURE; discrete = TRUE; states = (Easy, Difficult); parents = (Category, Weather); probs = // Easy Difficult // Category Weather (((0.5, 0.5), // Mentally Challen good (0.1, 0.9)), // Mentally Challen bad ((0.5, 0.5), // Despondent good (0.1, 0.9)), // Despondent bad ((0.9, 0.1), // SmChild good (0.75, 0.25)), // SmChild bad ((0.65, 0.35), // Child good (0.1, 0.9)), // Child bad ((0.75, 0.25), // MiscAdult good (0.5, 0.5)), // MiscAdult bad ((0.59, 0.41), // Elderly good (0.1, 0.9)), // Elderly bad ((0.75, 0.25), // Hiker good (0.67, 0.33)), // Hiker bad ((0.83, 0.17), // Hunter good (0.66, 0.34))); // Hunter bad ; comment = "For each category, Syrotuck estimates detectability\n\ in both good and bad weather. He gives percentages\n\ for \"Easily detected\" in both.\n\n\ Despondents and mentally-challenged are noted to be\n\ hard to detect -- they don't respond to searchers and may\n\ evade. I basically modeled them on ELDERLY, who\n\ also don't respond, guessing that evading was about as\n\ hard to spot as someone lying prone.\n\ "; whenchanged = 1020050751; visual V2 { center = (474, 342); height = 6; }; }; };