Geographic Profile

[wesleyenglish]

I, for one, would be interested in seeing a blended heat map based on C1, C2, C4 & C5.

Here you go! Any reason for leaving out C3 (Elizabeth Stride)?



Click on the image for the full size version.

[The Good Michael]

Here you go! Any reason for leaving out C3 (Elizabeth Stride)?

I say,"No." Others might tell you differently.

Mike

[corey123]

There are many descrepencies that indicate she may not have been killed by Jack the Ripper.

Check my thread named "Did Jack kill Liz Stride?"

[The Good Michael]

There are many descrepencies that indicate she may not have been killed by Jack the Ripper.

But none of those 'differences', not discrepancies, are conclusive enough to omit her... at this time.


Mike

[corey123]

Hi Micheal,

I don't omit her.

[Sam Flynn]

But none of those 'differences', not discrepancies, are conclusive enough to omit her... at this time.

The way I see it, Mike, the similarities between the others - viz., abdominal and other mutilations, over and above a cut throat - might make them a somewhat safer bet when it comes to profiling exercises. It's not so much a "ruling out", as an "erring on the side of caution", in my view.

[FrankO]

Here you go! Any reason for leaving out C3 (Elizabeth Stride)?

Hi Wesley,

Thanks for putting up the map without Stride!

As I’m unsure of Stride being a Ripper victim, I was curious as to how she influences the geographic profile. And I agree with Gareth (Sam) that the similarities between the others might make them a somewhat safer bet when it comes to profiling exercises.

All the best,
Frank

[corey123]

Hi Wesley,

Also notice that if you exclude Stride, all the murders are north of Whitechapel road and Commerical road.

If you remember Canters work, think Adrian Babb. If you remember this case, then you will get my meaning.


Yours truly

[Chris]

I almost feel I shouldn't ask, but - what is the actual model you're using to work out the spatial probability distribution? Where do the parameters come from?

[wesleyenglish]

The GeoProfile Model

Chris, no need to worry. I developed GeoProfile as a part of my thesis research at the Chicago School of Professional Psychology where I earned a master's in forensic psychology. Consequently, the GeoProfile model is no secret. It is detailed in my thesis, which can be found in its entirety here.

It's difficult to fully explain the model in any economy of space without an understanding of geographic profiling. If you're interested in the subject, I highly recommend taking a look at my thesis; I think it serves as a good, in-depth introduction to geographic profiling.

Basically, a grid is overlaid the search area and every cell is evaluated for its likelihood of holding the offender's anchor point (such as his residence or workplace). The likelihood score for a cell is determined by summing the distances to each of the crime scenes as adjusted by the distance decay function.

For this particular profile, I used a lognormal distance decay function, which takes the following form:



where dij is the distance between the cell and the current crime scene under evaluation, MeanD = .125 miles, Sd = 1.5 miles, and alpha (an arbitrary constant) = 1.

A distance decay function (DDF) is just a mathematical representation of the observation that offenders tend to commit crimes close to home with the probability of offending decreasing with distance from home. The lognormal DDF also incorporates a buffer zone, which is the idea that an offender will avoid committing crime too close to home to avoid detection and capture.

My model is based most closely on CrimeStat's journey-to-crime module, but also closely resembles the model used in Dr. David Canter's geographic profiling system called Dragnet. All three systems (GeoProfile, CrimeStat and Dragnet) use very similar math. The model used by D. Kim Rossmo in his system, called Rigel, is based on the same ideas, but uses different math to model them.

I hope that answers your question! As a social scientist, I am dedicated to transparency. My system isn't designed to corner the market; it's meant to advanced the field and increase the use of geographic profiling.

[Chris]

Thanks for those details.

What I always worry about with these models is the choice of parameters. Won't your distribution be quite different if you use different values of MeanD and Sd?

[The Good Michael]

The way I see it, Mike, the similarities between the others - viz., abdominal and other mutilations, over and above a cut throat - might make them a somewhat safer bet when it comes to profiling exercises. It's not so much a "ruling out", as an "erring on the side of caution", in my view.

I think erring on the side of caution goes both ways... and indeed, in the direction of Tabram as well. To be sure, Tabram doesn't change much being in the center of things. Still, two models would be nice. And good to see you back!

Mike

[tnb]

Fascinating stuff Wesley - I have made no secret of my slight cynicism towards the use of profiling per se in the investigation of crimes commited in a markedly different social time, but your summary is to my mind clear, logical and open enough to overcome some of those doubts. If it is okay with you I am going to give your work a mention and a link, with a more indepth analysis and discussion, on my website when I get a chance to update later today.

I would be very interested to know which other cases you plan on using for your 'infamy' series? From a UK point of view I am guessing Peter Sutcliffe is the ultimate 'commuter', taking on from your point about motorway exits it would be interesting to know whether your model could have if not isolated his home address for obvious reasons at least predicted some of the roads he used most regularly. More recently, the Stephen Wright case (the 'Ipswich Ripper') would perhaps be the closest I can come up with in terms of a closely packed distribution of crimes, also in a tight timeframe. Was GP used in that case, I cannot recall at this minute?

I would like to add my voice to the approval for including the map minus Stride - I am also glad I didn't have to be the one to suggest it! I notice however that taking her out does not appear to make a great deal of difference (albeit a slight creeping further eastwards towards Bethnal Green). As a beginner with these kind of tools I was a little surprised by this - in your experience is it normal for one crime to make so little difference to the whole construction? Would excluding any other victim make a more marked difference, for example - dare I say it - Kelly (5)?

Once again, thank you for a fascinating insight; I would appreciate any replies to the points I raised, even if you may find them a little naive. From a personal opinion point of view I have always felt the key to this crime (if it is ever to be found) lies closer to Whitechapel rather than Spitalfields, and I am pleased to see anything which appears to add credence to this! Dare I raise the ghost of Stephenson by pointing out that the Royal London Hospital lies pretty much bang in the middle of the 'red zone'...probably not.

Corey - I have read about the Babb case and Canter's work with it, although probably not in as much detail as you or Wesley admittedly, but I must admit the relevance to a non-Stride distribution being north of the Whitechapel and Commercial roads evades me. Would you be so kind as to enlighten me?

[corey123]

Hi Tnb,

If you have "criminal shadows", read pages 106-119.

Yours truly

[Sam Flynn]

I think erring on the side of caution goes both ways...

I can't quite agree with you there, Mike - at least not where there are clear features that set the disembowelling murders apart from a "mere" cut throat or multiple stab-wounds. I do, however, agree with you on the need for more than one map... with or without Swede, so to speak.

And good to see you back!

Thanks for the welcome, and it's nice to be back in your company, sir!