Oh, and I should say about the above routines, both work far better than random chance as well. Performance declines as the number of offenses declines, in the sense that as the number of offenses decreases the estimated "territory size" gets bigger. Generally, the crime zone area is calculated as the area of the smallest circle that encloses all of the offense locations - then the territory size is estimated by inflating this based upon the number of offenses, with fewer offenses resulting in a greater inflation factor. Then, the spatial routines produce a jeopardy map by calculating a priority value at each pixel of the inputted map. These are then arranged from highest to lowest, and grouped into "zones", with each zone comprising 2.5% of the estimated territory size. And you map out the zones, in this case stopping when you reach the number of zones required to reach the 75% probability criterion.
Although the above would indicate that Dr. Watson requires fewer zones, Dragnet the next, and Rigel the most, I'm not convinced those values are stable. I've got some analyses I have to do when I get an appropriate amount of time to devote to this, and my suspicions are that in the end the cut off zones will end up being the same for all three routines. While it would be nice for my ego if my routines out performed the others, I suspect they will end up being similar to the others. One of the major difficulties, actually, is getting a sufficient number of reliable maps to work with that don't come from a biased selection set. Maps included in the literature tend to be "exemplars", meaning, a map that work really really well for a given routine, so when one includes those in a test set, you bias the test set towards the routine that the article favors. (Note, this isn't deceptive on the articles part, as the demonstration is presented as just that, an exemplar of how well the routines can work; but it is an issue I have when trying to test the efficacy of the routines as I want the good, the bad, and the ugly!).
- Jeff
Announcement
Collapse
No announcement yet.
Open Mind
Collapse
X
-
Originally posted by Darryl Kenyon View PostHi Jeff
Would it be possible to do a map with the C5 plus Martha but minus Kate out of said C5 ? The reason being I am one of those that believe the killer had a second bolt hole near Mitre Sq which he used on that one particular murder. Hence the apron not being discovered in Goulston street until the later time.
Regards Darryl
Sure. That would look like this:
The Rigel routines include a "buffer-zone" in the calculations, meaning the offender is thought to be "close" to offense locations but not "too close". It also bases distances on "Manhattan distances", so you total the horizontal and vertical distances between two points (as the cab drives) rather than the "Euclidian distance", which is the distance in a straight line (as the crow flies). This is why the Rigel outputs tend to have sharp corners and such, as the underlying geometry is based upon "squares" rather than "circles".
Other routines, like those in "Dragnet" (by Canter), do not include a buffer zone (hence the highlighted areas right around offenses; this doesn't always happen, but does sometimes, as in this example), and tend to be based upon Euclidian distances. For example, if I do the above using Dragnet routines, we would get this, which produces a more rounded looking output due to the underlying geometry is based upon circles:
And my own routines ("Dr. Watson"), also employ Euclidian distances, but the analysis is based upon a number of different underlying patterns, which then get combined. One pattern is a general "distance decay from the offenses", but uses a different calculation from Dragnet, and because it involves other spatial aspects, it doesn't tend to produce "spikes" at individual crime locations .
All three sets of routines work equally well, meaning if I give them all a bunch of different cases, on the whole they produce similar results over the entire set. Obviously, for any individual case, one routine may work better than another, but that is neither here nor there when evaluating the generality of the routines.
And while at first glance the areas look different, in the end, they are all highlighting very similar regions really. At some point I want to try and take all three routines and combine their output to see if that in anyway improves performance. It's like getting three opinions, and looking for where those opinions agree. Whether or not that improves things is an empirical question though, and at the moment I've not started on that.
- Jeff
- Likes 1
Leave a comment:
-
Originally posted by JeffHamm View PostHi George,
Sure, I've had to use a different map in order to zoom out enough to include Coles, but not too much changes.
Thank you for that additional map. The significant change that I notice is that McKenzie and Coles extend the hot spot area to Middlesex St and the Butcher's Row area, which was of particular interest to Inspector Robert Sagar. This additional area is not added by the addition of Tabram, and that additional area on the C5 + Mckenzie map disappears with the addition of Tabram. Can this be interpreted that McKenzie and Coles were by the same hand, but that hand may not have been the same hand as for the C5 and Tabram, or perhaps that Coles is the outlier in the sample?
As we have discussed before, we have the choice of attributing all the Whitechapel murders to the same hand, or admitting the possibility that there was more than one killer involved. I am inclined to attribute the stabbing murders, Smith, Millwood and Tabram, to a different hand than the C5 + Mckenzie + Coles, but that is just my opinion at this point in time.
Best regards, George
Leave a comment:
-
Originally posted by Darryl Kenyon View PostHi Jeff
Would it be possible to do a map with the C5 plus Martha but minus Kate out of said C5 ? The reason being I am one of those that believe the killer had a second bolt hole near Mitre Sq which he used on that one particular murder. Hence the apron not being discovered in Goulston street until the later time.
Regards Darryl
It's interesting that the killer's most direct route to Goulston St is via Stoney Lane, right past Jacob Levy's home on the corner of Stoney Lane and Middlesex St.
Cheers, George
- Likes 1
Leave a comment:
-
Hi Jeff
Would it be possible to do a map with the C5 plus Martha but minus Kate out of said C5 ? The reason being I am one of those that believe the killer had a second bolt hole near Mitre Sq which he used on that one particular murder. Hence the apron not being discovered in Goulston street until the later time.
Regards Darryl
Leave a comment:
-
Hi George,
Sure, I've had to use a different map in order to zoom out enough to include Coles, but not too much changes. This is what makes me wonder if the spatial analysis is possibly, at least in part, sensitive to common spatial activities of the victims. That wouldn't make the information entirely useless, but it would change how one uses it. If this were the case, then the hot spot may not reflect JtR's anchor points (like residence, work, etc), but would be good places to do stake outs, looking for men not from the immediate area. No doubt JtR was on the hunt more frequently than he killed, so taking note of anyone not from the immediate area who appears regularly would be one idea. The spatial analysis in this way suggests areas where to set up those observation posts.
You may note, I use the term spatial analysis rather than "geographical profile", and that's because the latter is really just a PR term, while spatial analysis describes what is being done. Moreover, this sort of thing is not "evidence", anymore than creating a "personal space analysis" (i.e. hot spot = spouse, ex-spouse, partner, ex-partner; next highest = family members, friends, etc) is evidence. Just because a spouse is the most likely perpetrator isn't evidence that "This spouse" is the perpetrator. Rather, the analysis is about prioritizing searches. Personal space analysis creates a list, and one starts there and works down, eliminating people until someone cannot be eliminated, and following the list order speeds up finding the perpetrator in general. Spatial analysis, instead of telling you "who to look at" to search for evidence, tells you "where to look" to search for evidence.
- JeffLast edited by JeffHamm; 11-21-2023, 06:52 PM.
- Likes 1
Leave a comment:
-
Hi Jeff,
Thank you for the analysis maps. Could I beg your indulgence and ask to see the map that includes the C5 + McKenzie + Coles please?
Best regards, George
- Likes 1
Leave a comment:
-
Originally posted by The Rookie Detective View Post
thought please kind people?
Clearly, they did not have the benefit of studies of serial killers.
I reckon there is sufficient in the mode of attack and the injuries to render Alice a likely victim of the WM.
- Likes 1
Leave a comment:
-
Originally posted by GBinOz View Post
Hi RD,
The geographic profilers generally use the location of the bodies to produce a prediction as to where the killer may have lived, or where he may have established a base. I have seen a comment that the inclusion of Tabram does not alter the map very much at all, and I would say that the inclusion of McKenzie would also not produce a major difference.
...
The Ten Bells is about 0.3 miles from the murder site, The Brittania and The Queens Head about 0.2 miles. Could Jack have known these women from having drunk in the same Pubs, and stalked them? This might explain why they would have been at ease with him. Pure conjecture I know, but I find it more than coincidental that all the victims lived so close to the profiled area for Jack's residence.
Phillips excluded McKenzie purely on the medical evidence, but he initially excluded Eddowes on the same basis. Bond included McKenzie on the M.O., and on that basis I am inclined to agree. JMO.
Cheers, George
If I include Martha Tabram to the C5, we get this:
If instead I include Alice McKenzie to the C5 we get:
and if I add both, we get:
So your instincts are correct, there isn't much of a change as a result of adding either or both to the pattern produced by the C5.
One thing, however, is that these kinds of spatial analyses are thought to reflect underlying decisions being made by the offender. However, that need not be the case here. In the JtR series, we do have to keep in mind that the victims behaviour and choice of locations are not as independent as those of a serial killer who picks people at random throughout a city. In this case, many of the victims will be frequenting pubs, as you mentioned, or if soliciting, then probably going to similar areas. As a result, the spatial analysis could very well be picking up common spatial patterns of the victims, rather than something about JtR per se.
Moreover, there are other routines (equations), that are at least as successful as Rigel in narrowing down and prioritizing the search area. For the C5, they tend to also produce a peak area in a similar vicinity as the Rigel one, but shifted N slightly (running between Miller's Court and 29 Hanbury Street).
Anyway, that being said, having looked at these a few times, if the spatial analysis is picking up on JtR's choices as well, I also think the pubs up in that general area would be a good place to start looking.
- Jeff
- Likes 5
Leave a comment:
-
Originally posted by The Rookie Detective View PostI know that by considering Kelly as not being the last victim, it is inconvenient for those who favour those suspects who were dead by the time McKenzie was killed, but I think she's just a victim of bad timing and because the Kelly murder stole the limelight so to speak, anyone killed after Kelly never stood a chance of being taken seriously as an authentic victim of JTR.
thought please kind people?
The geographic profilers generally use the location of the bodies to produce a prediction as to where the killer may have lived, or where he may have established a base. I have seen a comment that the inclusion of Tabram does not alter the map very much at all, and I would say that the inclusion of McKenzie would also not produce a major difference.
The red zone in these maps centres around the Thrawl/Flower and Dean St area, extending up to the Heneage St area. Around the time of the murders, the canonical five all lived in Thrawl St, Flower and Dean St and Dorset St. McKenzie lived in Gun St, at the western end of Dorset St, not far from Miller's Court, and was murdered in Castle Alley ( the green dot).
Accordingly, one of the things the victims and Jack may have had in common was their patronage of the same Pubs - The Ten Bells, The Brittania (Ringer's) and The Queens Head. On the night of McKenzie's murder there was a report from a barkeeper:
At 12.30 when all the public houses are closed by law, the barkeeper of a "pub" situated a quarter of a mile from the scene of the murder, says that he turned her into the street, and that she had been drinking some, but was not actually drunk. making her way home the woman turned into Commercial street - the exact region where most of the other murders had been committed. Here all traces of her were lost till the body was found in Castle alley at 12.50 this morning.
The Ten Bells is about 0.3 miles from the murder site, The Brittania and The Queens Head about 0.2 miles. Could Jack have known these women from having drunk in the same Pubs, and stalked them? This might explain why they would have been at ease with him. Pure conjecture I know, but I find it more than coincidental that all the victims lived so close to the profiled area for Jack's residence.
Phillips excluded McKenzie purely on the medical evidence, but he initially excluded Eddowes on the same basis. Bond included McKenzie on the M.O., and on that basis I am inclined to agree. JMO.
Cheers, George
- Likes 2
Leave a comment:
-
While I believe that Alice may very well have not been a JtR victim but most likely was one, I do think that timing is a consideration when assessing the likelihood of a victim being a Ripper victim. But in Stride's case, it's not just that she was killed before Kelly, it's also that she was killed a little less than an hour before Eddowes. To me it seems like rather unlikely that there would have been 2 different men slitting the throats of prostitutes within an hour of each other in 2 locations that are within easy walking distance of one another.
On the other hand, when Alice was killed, some thought she was a Ripper victim and some thought that she wasn't, but of those who thought that she wasn't, I believe that their reasons were differences in her wounds, not that because 7 months had passed since the last murder, he couldn't have struck again.
- Likes 3
Leave a comment:
-
Originally posted by The Rookie Detective View PostIf Alice McKenzie would have been murdered BEFORE Kelly, then would she be taken more seriously as a ripper victim?
If for example Stride had been murdered after Kelly, then she wouldn't have even been considered at all as a possible JTR victim.
There are clear similarities to the C5 - mutilations and the body posed. OTOH, the level of mutilations is much less and the time gap between murders is much larger. Dr Phillips, who conducted or attended the autopsies of Chapman, Stride, Eddowes, and Kellly, did not think that Mackenzie was killed by the Ripper.
If she was killed by the Ripper, the lesser mutilations could be a sign of failing health on the part of the Ripper. Or perhaps the thrill was gone. Or McKenzie was killed by a copycat. I lean towards the latter.
- Likes 1
Leave a comment:
Leave a comment: