Stats on shootings are informative about past events, but there are too many variables for those stats to have any value as a prediction about your future gunfight.Originally Posted by RevolverRob
As far as outliers go, if I'm prepared for the outlier (more BGs and rounds) I'm definitely prepared for the norm (1 BG and a 2 rounds).
"Gunfighting is a thinking man's game. So we might want to bring thinking back into it."-MDFA
Beware of my temper, and the dog that I've found...
Rob,
I do not see a predictive analysis. What I see is a descriptive analysis. (My last analytics class was 25+ years ago and I may be misremembering my terms)
Therein lies an issue. The information presented is not predictive but is often interpreted as such even if it is not presented that way.
I think the information presented does have value, just to as much as some people may be inclined to give it.
One man's opinion worth exactly what you paid for it.
Last edited by Kevin B.; 11-29-2016 at 07:15 PM.
C Class shooter.
The standard deviation - useful for symmetrical, normalish distributions. Since the shots fired by civilians distribution is probably a J curve, starting at zero and tailing off to the higher numbers - is the SD worth it? If you discount the no shot incidents, you start the J at one and then descend.
You might consider a bivariate plot of number of attackers by shots fired also.
I've heard some variation on this line of thought expressed numerous times.
While I understand the rationale, this is often not well understood and can be incredibly confusing for less experienced shooters.
Being prepared for outliers does not necessarily mean one is prepared for the norm.
Fair enough. What I view in Claude's data and others is driven by my analytical nature. They have not, nor did they intend to, build a predictive model. However, the quantification of data in this way, allows you to assess and paramaterize exactly what is necessary to build a predictive model.
In this case you can boil your variables down as Claude did in the TAC material:
1) Place in which an attack is likely to take place: Home, Business, Public, Other - Based on percentages of incidents they are most likely to occur at Home and Business. Therefore, each of those parameters need to be weighted more heavily.
2) Gun Usage in Attacks - Zero Shots, 1-5 shots, 5-10 shots, 10 or more shots - Each can be quantified in the number of shooting incidents - And therefore can be used as a predictive parameter. Most appropriately, you would break these data down by location as well. E.G. are you more likely to fire more rounds in the home, business, or public?
3) Timing. - A critical component of this is timing. As in what hour of the day did it occur?
Just those three factors can allow you to build a predictive model and give you a set of probabilities for a given set of circumstances.
I have not done that with this particular data, but I did similar analyses investigating the FBI's "Mass Shooting" data last year. There - we discovered via randomly sampling that the strongest correlation between reduced number of causalities in mass shootings was intervention of any kind during the active shooting event. - However, even the number of variables simplified failed to capture all of the complexity of each shooting. In other words, even though intervention reduces casualty rate, it does so in different ways. Sometimes attackers are killed by their potential victims, other times they are disarmed, other times they see the resistance and give up, other times they see the resistance and kill themselves.
The model can't tell you what an attacker will do, but it can tell you that confronting an attacker reduces casualties in mass shootings. It also doesn't assume that intervention WILL reduce casualties in a given scenario, it simply tells you that given an attacker and intervention, intervention usually reduces the number of casualties.
___
Similarly data of this type can inform you about a variety of things. When, where, number of potential attackers, and number of potential rounds of ammunition needed to resolve a fight. However, without actually assessing each of the variables in a predictive model, it cannot tell you what you "should" do. BUT I maintain that said approach can be used to build an informative model for YOUR LIFE.
For instance, when I moved to Chicago. I was determined to develop a model that could predict the non-random, non-normal distribution of violent crime in the city. Though I haven't been fully successful yet, what I have managed to identify is a key set of parameters necessary to predict a violent encounter. And shock of shocks, it revolves around, Location, Time of Day, Activity of Persons in the incident. And allows you to predict, with alarming, accuracy when and where violence will occur based on past incidents of violence and the given parameters.
All of that is my long-winded way of saying...Data is good, but can be skewed and is often misinterpreted. People like to take things like, "Mean number of rounds fired in a gunfight." And say, "Great, the mean is 5, so if I carry 10, I'll be prepared." - That focuses on one of a dozen potentially important factors of a lethal force encounter. However, many of those factors are less "fun" than number of rounds fired in a gunfight. And instead focus on things like, when, where, and who were involved in those fights and how were they ultimately solved.
Also, Glenn is a 100% on these types of data not being normally distributed. When I did the FBI Mass Shooting stuff last year, the very first thing I did was test for normality in my various factors and found none of them to be normally distributed. Which means we can either transform the data to be normally distributed or use non-parametric statistical analyses, but this is something that is often overlooked. There is frequently the assumption of normality which is probably not the case.
Last edited by RevolverRob; 11-29-2016 at 09:32 PM.
"1) Place in which an attack is likely to take place: Home, Business, Public, Other - Based on percentages of incidents they are most likely to occur at Home and Business. Therefore, each of those parameters need to be weighted more heavily."
Not so. In the NRA reports, the defender is at home or in his place of business because that is where he has access to a gun with which to defend himself. That does not mean the attacks themselves happen more frequently there. Since most people don't carry away from home/business, that skews this data set. Actually, only 13% of armed robberies, for instance, occur in or around the home. Most occur in public places. That holds true for most violent crime.
This jives with everything detectives I've spoken to have to say on the issue as well.
Tom, is it accurate to say that a decent chunk of those armed home invasions involve victims that are linked to drug activity or living in homes previously linked to drug activity? This is a common thread mentioned to me by multiple local investigators; that at least around here a good portion of armed invasions are a result of the perps having reason to believe there are drugs or illicit money in the home.
That used to be the case. Now, we are seeing more and more where the victim simply lives in a decent house, that looks like there might be valuables present.
I live in a very decent suburban community and we've had a few home invasions occur nearby. One involved gunfire -- the homeowner fired shots but the intruder escaped the house. I was out on my back deck (armed) watching over our back hill since the house is directly down behind ours but I immediately went back inside when a police helicopter appeared overhead with a very bright searchlight. The suspect was caught that evening.
Extremely sobering. I practice something like parking my locked next to an unlocked Porsche - my house is the least attractive on my block.
So much this. Preparing for outliers seems like a terrible idea if it interferes with or replaces your preparations for the norm.
The answer, it seems to me, is wrath. The mind cannot foresee its own advance. --FA Hayek Specialization is for insects.
Reply With Quote