Instructions for “Probability Mapper”

NOTE: This is OLD software. It was only a proof-of-concept. It was NEVER operational.  Needless to say, it is completely without warranty, etc.  Installation

 1. Download "Setup.exe" to your desktop.

2. Unzip it to C:\PM.  This is IMPORTANT.  The demo files have "hard coded" filenames and will not work from anywhere else.  When you are done, you should have the following files.

Note the directory: C:\PM

Running PM

Probability mapper can be run directly, without talking to the resource allocation code or Access.  There are 4 demo files you can load and play with. We will work on one here.

Launch "PM.exe".  That should give you a blank window:

Choose File->Open and select "demo1.sar"

Your screen should now have an annotated map. A small portion of it:

This already has a subject profile. Select Case->Edit case details to see that it's a 3yo boy.  There are some extraneous fields showing too.

The View menu has some interesting options:

Using the menu, turn off Distance Rings and turn on Stat Rings.  Now the screen has percentile rings for this 3yo child, (based on some very old statistics).

Change the subject to a 25yo Hiker.    Notice that the stat rings change to be much wider (about 3km out):

(Notice the unrealistically close 99% stat ring. Did I mention this was only a proof of concept?)

If we select View->Network Graphs, we get a (really low res) view of the distributions themselves:

Go back to View->Map Screen.  (Yes, it should be a toggle.)

Now, Select View->Shade Regions.  You will see the regions (squares) roughly shaded by probability.  (Adam wasn't able to get transparency working in his toolkit, or fine shadings. Those would be trivial now.)

Actually, as far as I can tell, these shadings don't change as we change profiles, so either the color profile is very course, or this feature broke in the packaged version. That's too bad, as it was the flagship feature of the software. I must look into this.

OK, turn off the shadings so we can see the map again.

If you right-click on the map, you get a context menu which lets you create, view, or edit regions.  It also lets you set the terrain/vegetation type, which changes the color of the region border. It should also alter the POA for the area, though the packaged version seems not to use them.


Another small bug: the list of terrain types repeats itself repeats itself repeats itself.

Selecting "Mark Location" lets you create a new labeled point, like this one:

Selecting "Draw Region" lets you draw a polygon region, which will get a POA based on distance (and in theory, terrain type).  Right click to set terrain type. Here we have drawn an irregular polygon and set the terrain type to "drainage":

Notice that it does not need to be connected to other search areas.

Loading a new behavior model

PM can use different behavior models. Just "Select Network File" from the case menu.


The demo comes with two behavior models, both early models from Adam Golding's thesis work, based on the limited Virginia dataset. The default is "honors2.dne". There is also "snobonly.dne". The SARBayes download page has links to some other models, including our 2002 version of Syrotuck's model. The statusbar reports which model you are using.


Descriptions and evaluations of these older models are available from the downloads page under "Other Reports". The most recent is my 2002 NASAR presentation is: http://sarbayes.org/nasar.pdf . The good news is that these models could predict the right distance (to within 1km) about half the time. The bad news is that they did just as well by disregarding lost person category. Now that we have a much larger database available (ISRID, the International SAR Incident Database), we hope to create and test some other models.

When we do, PM or its successor can just load them!

Other Map Functions


The first three options let you load, calibrate (georeference) a map, and set the PLS. The context menu lets you quickly draw an initial grid. The last two options let you save (dump) the POA map to a file, or load one from a file. (We called it POC then, sorry.) These are very useful for communicating with a resource allocation program, like SAR.exe (included).

Admittedly, communicating with a text-mode allocation program is cumbersome. It would be nice if they were integrated. They are, in a chewing-gum-and-baling-wire sort of way, using Microsoft Access.

 

Launching from Microsoft Access

This part requires Microsoft Access.  It was written in Access97, and may not run on later versions. However, you can download the runtime from: http://www.zlcsoftware.com/ftp/A97RT.EXE

André reports better luck using the Access97 viewer available from an EPA underground storage tank database website.

Direct link to the installer: http://www.epa.gov/swerust1/ustacces/runtime.exe

The EPA database website, with instructions: http://www.epa.gov/swerust1/ustacces/uav30.htm

Launch AGM97.mdb: . You should see a screen like this:

The top "Main Menu" window is the GUI, such as it is.  The other window shows the tables that were used to make it, and we will ignore it.

If you are lucky, you can now "Load a Map from Probability Mapper", use the default resources (or modify them using an Access GUI), and "Suggest" allocations with either Charnes Cooper or Washburn.  These suggestion routines will run SAR.exe with the appropriate calls.

However, I find that I can no longer Load, Import,  or Export anything, because it is an Access97 file running on my copy of Access2003.  If I use the version installed by A97RT.EXE, I get a different error about an ISAM file.  Perhaps more later.

 

Author: ctwardy

Charles Twardy started the SARBayes project at Monash University in 2000. Work at Monash included SORAL, the Australian Lost Person Behavior Study, AGM-SAR, and Probability Mapper. At George Mason University, he added the MapScore project and related work. More generally, he works on evidence and inference with a special interest in causal models, Bayesian networks, and Bayesian search theory, especially the analysis and prediction of lost person behavior. From 2011-2015, Charles led the DAGGRE & SciCast combinatorial prediction market projects at George Mason University, and has recently joined NTVI Federal as a data scientist supporting the Defense Suicide Prevention Office. Charles received a Dual Ph.D. in History & Philosophy of Science and Cognitive Science from Indiana University, followed by a postdoc in machine learning at Monash.

Leave a Reply

Your email address will not be published. Required fields are marked *