Geographic Profiling

[Abby Normal]

It sounds like you are thinking of Mrs. Fiddymont's suspect in the Prince Albert, 21 Brushfield Street, near the corner of Steward Street. Looking in the mirror, she noticed he had caked blood on his hands.

When the man was followed, he walked west towards Bishopsgate, and then turned south.

yup-and blood behind his ear too. I beleive he was described as reddish hair and fair/ruddy complexion?

[caz]

Hi Caz,

If JtR were aware of where the police boundaries were, and the implications of that, then yes, I can see an argument as to why he might have chosen that location for Eddowes following the Stride murder. Whether or not he was so aware is something we may never know, but it is an idea that has some interesting implications with regards to Eddowes' crime location.

The problem, though, is for us to accept that line of reasoning, we have to suggest that JtR was thinking pretty quickly on his feet to both decide he was going to commit another murder that night, and also to specifically target this impromptu murder in area within City Police territory. And to be confident he would find a victim in time. I would think doing that would require more specific planning, so could see an argument that it would be more likely if, in fact, Stride was not a JtR victim. (I can see it working either way, basically).

Regardless, it's an interesting idea to ponder, but what was actually going on inside JtR's mind at the time, even if we were to solve the case, is something that I think is lost forever.

- Jeff

Good points, Jeff. As you say, it can be argued either way. There had now been 3 weeks without a murder, so the killer may have been planning to try his luck over in the Aldgate area anyway, whether or not he knew about the police boundary between. It does strike me as coincidental, though, that on the very night he killed again, he found himself in the City, at a time which allowed for him to have walked there from Berner Street and found Eddowes along the way. I'm not sure he had to be confident of finding a second victim 'in time'. He was simply in the mood that night, but couldn't know if and when opportunity would knock. The opportunity to mutilate that night came when he was in Mitre Square with Eddowes. He could have called it a night at any point before that, if she hadn't come along when she did, regardless of whether he'd previously been in Berner Street. But that would help explain the change of direction, to a location that was not in the vicinity of any of the previous murders, including Stride's. If two killers were acting independently, neither had an alibi for the other one's murder.

Love,

Caz
X

[jerryd]

Notice the date and time on this one.

St James Gazette
November 12, 1888

[Darryl Kenyon]

I believe that PC Long was telling the truth regarding the apron. In other words I feel Jack had a bolt hole somewhere in/near Aldgate, possibly a place of work. [ Robert Sagar ]. So after a failed attempt to mutilate a woman in a first comfort zone he moved to a second before heading home. In essence that's why I wanted to see what a Geo profile threw up IE without Kate but with Martha and possibly Emma which Jeff has kindly shown.
Regards Darryl

[Scott Nelson]

Every time a victim is added or deleted, the boundaries change significantly. There were too few victims (data points) to define any meaningful trends.

[JeffHamm]

It sounds like you are thinking of Mrs. Fiddymont's suspect in the Prince Albert, 21 Brushfield Street, near the corner of Steward Street. Looking in the mirror, she noticed he had caked blood on his hands.

When the man was followed, he walked west towards Bishopsgate, and then turned south.

Hi rj,

Yes, thank you. That's the story I'm thinking of. Brushfield Street is the next one north of Dorset.

- Jeff

[JeffHamm]

Every time a victim is added or deleted, the boundaries change significantly. There were too few victims (data points) to define any meaningful trends.

Hi Scott,

Don't pay too much attention to the boundaries below the orange areas, by that point, we're just dealing with values needed to do proper statistical analysis, for the application, all that is going to be of real concern is the highest priority zones, which would be out to the orange. Yes, they shift a bit, but what is fairly consistent is that the profile is focused in that North West area, so in the vicinity of Dorset and Commercial. That takes in Flower and Dean, and those sorts of areas, extends generally up towards Hanbury, and also south a bit towards and just past Wentworth. Basically, more or less the area the police were interested in 1888. Different routines estimate different areas, Rigel focuses on Flower & Dean more, etc, but basically, all of the different approaches hone in on that area. Five locations is low, and Rossmo suggets that is the minimum required, but I don't know of any systematic study that has tested that value; meaning why not 4, or 3, what's magic about 5?. Regardless, 5 is considered sufficient for these routines, and really, while the lower zone boundaries shift around, that's not as big a deal as the fact the high probability area is stable (it's only the high probability area that would be of interest to the police, or in this case, us).

Again, the area encompassed out to the end of the orange is the high probability zone of interest. And when I've analysed a lot of other cases, what one finds is that 50% of the offenders will be located in that zone. 75% are found inside the area that spreads out to the light red border, and 90% end up inside the red border. Sadly, the data set of offenses I have to work wtih mostly come from solved cases I've found maps for online (BTK, Ted Bundy, etc) or from books on Geographcal profiling (where the cases presented tend to be highly suited for this type of analysis). I have a set of serial arson cases, which were all of the serial arson cases in New Zealand over a period of years, and when I test just those, similar results are found, so that gives me some confidence the inclusion of the cases from the books are not skewing things to look better than they should. However, there is some evidence of that. Rigel, for example, does better on the cases presented in those books than Dragnet, while Dragnet does better on the cases presented in books about Dragnet! Hardly a surprise, since one is going to use good results to illustrate things, but a concern for people like me who want a bias free data set.


- Jeff

[DJA]

[Wolf Vanderlinden]

Every time a victim is added or deleted, the boundaries change significantly. There were too few victims (data points) to define any meaningful trends.

Yes. I was told that a minimum of 10 points was needed to even begin to suggest any accuracy. Therefore the geoprofiling of the Whitechapel murders was considered to be just "an interesting exercise" and nothing more.

Wolf.

[JeffHamm]

Yes. I was told that a minimum of 10 points was needed to even begin to suggest any accuracy. Therefore the geoprofiling of the Whitechapel murders was considered to be just "an interesting exercise" and nothing more.

Wolf.

Hi Wolf,

Not sure who told you that, but generally 5 locations is cited as the minimum in the research literature. There hasn't been a lot of work to empirically determine a minimum though, but in the analyses I've done (albeit on a limited data set), 4 or 5 seems about right. It does appear to get a little better as the numbers go up, and in the vicinity of 10 is probably where it asymptotes, but performance with as few as 4 cases is still very good (as in overall 50% fall in the hot spot area; that's the pink, yellow, and orange areas; and 75% when the light red area is included. That corresponds to roughly 10% (8.89%) and 20% of the total search space; (I think I said 7.5% before, it's 8.89% for my own routines, which I round up to 10% to be conservative) for 50% or 75% return. If I limit it to only cases with 10 or more, 50% fall in 7% of the search area, and 75% in 15%. So, there's a slight improvement for the 50% cut off, and the 75% cutoff tightens up even more. However, I'm a bit cautious about those values as I don't really have a big enough sample to be confident those are stable yet (there's only 48 different series with 10 or more; 100 have 4 or more). Hmmm, just did a check since those values for 4+ include the 10+ series as well. If I limit it to between 4-9 (so drop 10+), the 50% area doesn't change (around 9% of the search space), though the 75% increases to 25% of the search space. Basically, it's the series that fall further out that get worse, which is probably not surprising as that suggests there is something a bit "unlike the others" in their patterns, so it takes more information to work that out.

But, one still has to remember, that's how it does when you submit a bunch of different series. For any individual case, like this one, that means there's a 50% or 25% JtR is not in those areas. I wouldn't dismiss any suspect where there was evidence that pointed towards them simply because they lived outside of the area indicated. It can be useful as a way of deciding where one might look, or to rank otherwise equal suspects (i.e. if there are two suspects being debated, and both have arguably similar supporting evidence pointing towards them, and one is in the suggested area, and one is not, odds favour the former; but those odds mean nothing if the latter can come up with more direct evidence).

These are just probability distributions, they can be useful when used correctly and not over hyped. Unfortunately, what tends to be presented in the public domain through media and/or popular films, is a gross distortion of what they do.

- Jeff

[DJA]

Yes. I was told that a minimum of 10 points was needed to even begin to suggest any accuracy. Therefore the geoprofiling of the Whitechapel murders was considered to be just "an interesting exercise" and nothing more.

Wolf.

Meh,there's a sucker born every day.

[Wolf Vanderlinden]

Not sure who told you that

Well, actually, it was this guy from Saskatoon named Kim Rossmo. I contacted him when I wrote The Murder of Martha Tabram: No Ordinary East End Crime back in 2006. He helped me with the geo profiling info for the article.

Wolf.

[JeffHamm]

Well, actually, it was this guy from Saskatoon named Kim Rossmo. I contacted him when I wrote The Murder of Martha Tabram: No Ordinary East End Crime back in 2006. He helped me with the geo profiling info for the article.

Wolf.

Hi Wolf,

Really? It's his books and papers I was thinking of as he has generally recommended a minimum of 5 in published articles. I'm surprised to hear that his view has changed quite so dramatically. I've not been able to find a decent empirical work published on the subject, though it's been a while since I've specifically searched so I'll have to do another literature search to see if he's got something out. Mind you, if he said that back in 2006 I would think it should be out long before now and I would have come across it (not necessarily though, depending upon where he's published it, if he has). As I say, I've not found any major drop in accuracy for less than 10 (specifically between 4 and 9), though there is some, but my data set is limited and not ideal, which could very well be why that's the case. Knowing researchers, personal opinions in conversations can be quite different from what they publish, either becoming more or less conservative. But going from 5 to 10 is a pretty huge shift.

Anyway, thanks for that.

- Jeff

[DJA]

RIPPER NOTES The Hunt for Jack the Ripper Issue No. 25 by Norder (Dan) Ed. by:: As New Softcover / Paperback (2006) First edition. | Loretta Lay Books

[DJA]

Environmental Criminology Research Inc. - Geographic Profiling - Crime Analysis » Training