@freemo @trinsec

At the risk of diving into statistical weeds deeper than my understanding, there is a strong positive correlation between gun ownership and homicide rates:

https://phys.org/news/2022-05-gun-ownership-homicide-stronger.html

https://www.hsph.harvard.edu/hicrc/firearms-research/guns-and-death/

Returning to your original post, there were 745 US deaths in mass shootings in 2021. Only 11 Americans were killed by lightning that year. (By odd coincidence, that was a record high year for gun deaths and a record low year for lightning deaths. There were a total of 444 lightning deaths from 2006 through 2021, an average of 27.75 per year.)

To say that lightning is more dangerous than mass shootings is wildly inaccurate. Comparing all PEOPLE killed by lightning to CHILDREN killed specifically in mass shootings AT PUBLIC SCHOOLS is cherry-picking.

It's a pretty safe bet that there will be multiple mass shootings in the US next week and there will not be multiple Americans killed by lightning.

@peterdrake @trinsec

> At the risk of diving into statistical weeds deeper than my understanding, there is a strong positive correlation between gun ownership and homicide rates:

I have spoken out on this as a professional data scientists many times. This sort of analysis is overwhelmingly dishonest for anyone with any training in data science who makes any attempt at being objective, and there is a reason for this.

That same correlation you show, guess what, youd see the same thing if you compared mental health access, or poverty levels, or any number of factors that lead to people being more or less healthy well rounded people. This is exactly why it is extreme intellectual dishonesty, you have no clue what of the countless factors are causing this, and as we all know correlation doesnt even **hint** at causality, it si corelation.

Now we **do** have something that is strongly suggestive of causality, they are called causality tests in statistics, and while they arent perfect anyone trying to be intellectually honest would use this method for decided how to analyze the situation. The simplest of these is called granger causality. When you do this form of analysis you more or less rule out other confounding factors by looking at the change in one value as another value changes. For example we look at.. does homicide rates go up when guns are banned, and vice versa.. you look for one event **following** another.

And guess what, when we provide these intellecually honest methods tot he data we see a very clear trend where banning guns tend to make violent acts and rapes skyrocket, and reversing bans on guns tend to cause those rates to drop (see some examples attached)

> Returning to your original post, there were 745 US deaths in mass shootings in 2021. Only 11 Americans were killed by lightning that year. (By odd coincidence, that was a record high year for gun deaths and a record low year for lightning deaths. There were a total of 444 lightning deaths from 2006 through 2021, an average of 27.75 per year.)

Why did you just run odds on something different than what I stated? No one was talking about mass shootings or lightening deaths, nor do those numbers have relevance to the point being made, which was that using school shootings as a political point to create action is pointless since they are exceedingly rare... I never argued mass killings (by gun or any other means) was something we shouldnt address

> To say that lightning is more dangerous than mass shootings is wildly inaccurate. Comparing all PEOPLE killed by lightning to CHILDREN killed specifically in mass shootings AT PUBLIC SCHOOLS is cherry-picking.

At no point did I, or anyone, make this claim. I never said lighting was more dangerous, I said it was more common to be struck by lightening, again, to make the point about how rare school shootings are, not to depict the danger of lightening.

@freemo @trinsec

First, on the question of correlation vs causality, my understanding is that there are four possible explanations for a strong correlation between A and B:

1) It's a coincidence. Significance tests and replication can rule this out.
2) A causes B.
3) B causes A.
4) A and B are both caused by a third factor.

Does this match your understanding? If so, which one do you think is at play here (where A is high gun ownership rate and B is high homicide rate)?

Your one example does show a spike a few years later, but it disappears immediately. What we ideally need is a control country that is like Ireland in every other way but didn't confiscate guns in 1972. Without this policy, aside from the one-year spike, would homicide rates increase, stay steady, or decrease?

Now, on to questions of misattribution.

Your specific original claim was:

--
In other words a child is more than **twice** as likely in the USA to get struck by lightening as they are to die in a school shooting.

Should we still mourne and be outraged by it... sure.. does that mean it is a problem that is common enough to be a huge concern... not really. We should probably put more effort into addressing the "lightening problem" than we should be about addressing school shootings.
--

(BTW, is there an easy way to quote text with >?)

I read two things into this. Let me know if I've misinterpreted you.

1) The media overemphasizes school shootings specifically. Your child is extremely unlikely to die this way.

(I'll grant you this one; see also plane crashes and shark attacks.)

2) Because of these splashy mass shootings, the media makes too big a deal out of gun deaths. It would make more sense to worry about lightning, which is more likely to kill you.

It is to this second implication that I am reacting. That is why I brought out the general mass shooting numbers.

(I assume being hit by lightning is almost always fatal, so I didn't distinguish between being hit or killed.)

If you "never argued mass killings (by gun or any other means) was something we shouldnt address", I apologize for misreading that.

How do you suggest we address that problem?

@peterdrake

> 1) It’s a coincidence. Significance tests and replication can rule this out.
> 2) A causes B.
> 3) B causes A.
> 4) A and B are both caused by a third factor.

So.. this is sorta valid but a gross oversimplification. As a professional who tries to do this stuff quite often it might help if i explain how we map these things...

So A causing B is what we call a "conditional probability". In reality there isnt just "some third factor" but a huge map of conditional probabilities with A effecting B effecting C and D effects B as well, etc etc. So we have a MAP of probability. We call this a "graphical model". The job of a data scientist is to discover that graphical model, define the probabilities between the individual links, and from that model the events.

When your graphical model is incomplete and you have influencing factrs you havent mapped we call that "confounding", sorta.

> Your one example does show a spike a few years later, but it disappears immediately. What we ideally need is a control country that is like Ireland in every other way but didn’t confiscate guns in 1972. Without this policy, aside from the one-year spike, would homicide rates increase, stay steady, or decrease?

You are correct of course that I am giving somewhat shallow arguments here, largely because if i get too technical it will probably be too complex for the scope here When I discuss this issue with fellow data scientists our models are far more involved. I am trying more to show the simple to understand elements that are part of a much bigger and more rigorous line of research one would need to do.

> (BTW, is there an easy way to quote text with >?)

You just have to do it manually as far as i know :(

> 1) The media overemphasizes school shootings specifically. Your child is extremely unlikely to die this way.

Yes this was one of my main points

> 2) Because of these splashy mass shootings, the media makes too big a deal out of gun deaths. It would make more sense to worry about lightning, which is more likely to kill you.

That was more for dramatic effect than a literal point, no.. My point is more that both of these things are such exceedingly rare events they arent really worth being our focus.. Problems with overall violence, and if guns have a role worth considering in that, is the bigger issue.

> (I assume being hit by lightning is almost always fatal, so I didn’t distinguish between being hit or killed.)

90% of people struck by lightening survive actually.

> If you “never argued mass killings (by gun or any other means) was something we shouldnt address”, I apologize for misreading that.

Mass killings, and really any killings, are a huge problem that needs addressing... Guns just arent really the problem anymore than spoons are the problem when it comes to physical health in the USA... our problem is lack of access to mental health, and generally poor environment.

Ironically this problem is usually very much exemplified in most internet threads on gun violence where people are so hateful and toxic to eachother it gets vile, demonstrating the very mental-health issue driving the violence, far more than the guns.

> How do you suggest we address that problem?

1) is the hardest solution, its social.. we have to teach people to be caring to other people, something that as i point out, is usually at its most vile when we discuss these topics.

2) prioritize access to mental health.. im not talking asbout fixing healthcare, thats just aprt of it, but also normalizing mental health and psychiatry and many other aspects.

@trinsec