Free Allegiance Forums

Swizerland is in many ways socially conservative, and the ever present assault rifles are a big problem.

When the men get depressed/jealous/suicidal somebody dies, often several people. A lot of people are trying to change the system there because the number of shootings is unacceptably high.

It's far from a perfect model... but you are right it does sound rather like the 2nd amendment arrangement. Thos people really believe in the modern day militia. They even have anti-tank weapons at home.

I plot a random line or not so very line on that graph to and call the R anything I like if you so smartypants Adepticus care to explain to other dumb fools what regression analysis you decided to bestow on us and how did you handle the clear outlier Sats are so easy when you close your eyes call R = 0.5 and say that things are related! clearly people with 5 guns per 100 people will prevent homicides too!!!one1!

I can

Can't just throw the data out.
My guess is the author of the graph never checked for orthogonality with even the most basic tools at his disposal. It is a common mistake and the check is timing consuming as well as tedious if done right. If they did, and it was ok then "everything else" is convoluting the factors they used to obtain the equation. So you redo the matrix, add in other factors, and do it again. Sooner or later, assuming continued orthogonality, data is availible, and decent intuition (or usually good luck) in the factors driving the question you'll get there.


You calculate R, not pick it
0.5 is an awful fit, worse than 0.754. Both say toss it in the waste bin to me. No where do they identify the C but lets call it the traditional and industry popular 0.9. You now have something that isn't much better than a coin toss. The 0.5 R and 0.9 C say your approach is biasing things and making your results worse.

Linearity does not mean just Y=mx + b.
It is perfectly acceptable to have curves and such... matter of factor it tends to produce a better fit

Sooooo before Adept does some splaining, you might want to make sure you will understand it correctly if produced


tl/dr
The graph's author seems to have an agenda and is hoping that the average person doesn't understand the information being presented to them or the details that are missing means it is basically ass

MrChaos wrote:QUOTE (MrChaos @ Dec 21 2012, 06:41 AM) I can

Can't just throw the data out.
My guess is the author of the graph never checked for orthogonality with even the most basic tools at his disposal. It is a common mistake and the check is timing consuming as well as tedious if done right. If they did, and it was ok then "everything else" is convoluting the factors they used to obtain the equation. So you redo the matrix, add in other factors, and do it again. Sooner or later, assuming continued orthogonality, data is availible, and decent intuition (or usually good luck) in the factors driving the question you'll get there.


You calculate R, not pick it
0.5 is an awful fit, worse than 0.754. Both say toss it in the waste bin to me. No where do they identify the C but lets call it the traditional and industry popular 0.9. You now have something that isn't much better than a coin toss. The 0.5 R and 0.9 C say your approach is biasing things and making your results worse.

Linearity does not mean just Y=mx + b.
It is perfectly acceptable to have curves and such... matter of factor it tends to produce a better fit

Sooooo before Adept does some splaining, you might want to make sure you will understand it correctly if produced


tl/dr
The graph's author seems to have an agenda and is hoping that the average person doesn't understand the information being presented to them or the details that are missing means it is basically ass

Deal with outliers correctly you ass! Outliers like Adepts are bane of our society! Throwing them out is an acceptable social metric and should be done more frequently! That you noticed the pun of R=0.5 good of you is, missing the correlation vs causation case you are!

StinkerTod wrote:QUOTE (StinkerTod @ Dec 21 2012, 01:02 PM) how did you handle the clear outlier Sats are so easy when you close your eyes call R = 0.5 and say that things are related!

Are you kidding me? Something that is smack on the trend line like that is pretty much the opposite of an outlier.

Go back to school An extreme case =/= outlier.

If anything the graph is suspiciously neat, but I'd like to see you challenge the basic implication of more guns => more people getting shot.


Anyway. I didn't come here to feed the trolls but to share this for some food for thought.

http://en.wikipedia.org/wiki/List_of_count...ated_death_rate

Finland's place on that list is nothing to be proud of... but I suppose I can be happy about the fact that my countrymen mostly use their guns to shoot themselves rather than anybody else.

Can I also point out that this data seems a bit useless because it doesn't include information on homicides by other means? I mean if people are just using knives or poison instead of guns, does it make a difference?

TakingArms wrote:QUOTE (TakingArms @ Dec 21 2012, 10:56 PM) Can I also point out that this data seems a bit useless because it doesn't include information on homicides by other means? I mean if people are just using knives or poison instead of guns, does it make a difference?

The data, when available shows that aggression happens without guns, of course, but fatalities are much more rare.

StinkerTod wrote:QUOTE (StinkerTod @ Dec 21 2012, 09:47 AM) Deal with outliers correctly you ass! Outliers like Adepts are bane of our society! Throwing them out is an acceptable social metric and should be done more frequently! That you noticed the pun of R=0.5 good of you is, missing the correlation vs causation case you are!

Well just no to most everything you have typed. If iirc irl you deal with numbers a bit so maybe it is all for fun to see how many times we'll respond. ok I'll do it because I'm enjoying myself... man I can be quite the sad mother$#@!er huh

You cannot throw out the outliers because you think they are outliers without more justification then they are way off the curve fit. The most robust method is to use any of the slew of statistical tools associated with the job and check the under lying data for bias. You have to be very careful when doing it but it is just as likely Switzwerland is a bad data point as the US. That was the bit about orthogonality, checking that data is not being corrupted by other factors such as TA's point about knife death's or The Republic of Panda Humpers is grossly misreporting things or someone fat fingered the data are all great examples.

The fact people arbitrarly throw out data because it doesn't fit the curve is biasism most of the time.

Now it is acceptable to toss out data points that are outliers however you need compelling real world evidence to do it. The stronger the effect of throwing it out the stronger the case to do it that is needed. For example: In a lab envirnoment where you are taking measurements of bearings coming off a manufacturing machine if you have 300 samples all within +/- .0578mm except for one which was + .82546mm well just leave it in is my call but yeah in this case it probably is a mistake. If it was just 30 samples it needs stronger reasons. Such as it was measured at end of the shift before the XMas holiday, where the operator was found to be drunk at his machine then there is strong case for removing it.

Soooo there should be somee linkee provided showing the reasoning behind the little dots on the graph and all of the data including the @#(! that was tossed out and how it effected the results.

Even then you missed pointing out the burping 800lb Aardvark in the coat closest when trolling Adept:
Honestly the best question of all about this whole $#@!ing graph stuff is what was the question the graph was trying to answer in the first place.

See Leexal's post for the tl;dr version