Author Topic: Actual statistician crushes Molyneux.  (Read 4305 times)

0 Members and 1 Guest are viewing this topic.

: o )

  • FDR Wizard
  • *****
  • Posts: 589
  • Respect: +651
Actual statistician crushes Molyneux.
« on: February 16, 2016, 03:26:51 PM »
+9
An actual statistician (rather than Mike DeMarco + confirmation bias + laptop) pulls apart Molyneux's powerpoint politics.

https://www.youtube.com/watch?v=4uVPvbw-k18
"In every ideologically based community, there's a group that's elated the confirmation bias machine has finally gotten around to visiting their neighbourhood, like the ice cream truck" . . money detonator

QuestEon

  • Just some guy with a blog.
  • Administrator
  • FDR Wizard
  • *****
  • Posts: 860
  • What's your opinion? I'd love to hear it!
  • Respect: +459
    • FDR Liberated
Re: Actual statistician crushes Molyneux.
« Reply #1 on: February 16, 2016, 06:29:40 PM »
+4
Too good not to make the front page of blog....

An actual statistician schools Stefan Molyneux

It isn't about winning the debate. It's about the truth.

money detonator

  • FDR Wizard
  • *****
  • Posts: 717
  • Respect: +761
Re: Actual statistician crushes Molyneux.
« Reply #2 on: February 16, 2016, 06:37:35 PM »
0
I do want to correct him.  Molyneux was never a "Philosophy major"  ;D

I totally can see Molyneux blaming the errors on his "researchers" if ever confronted with this stuff.

Does regression abuse violate UPB?  ;D
« Last Edit: February 16, 2016, 08:12:34 PM by money detonator »

QuestEon

  • Just some guy with a blog.
  • Administrator
  • FDR Wizard
  • *****
  • Posts: 860
  • What's your opinion? I'd love to hear it!
  • Respect: +459
    • FDR Liberated
Re: Actual statistician crushes Molyneux.
« Reply #3 on: February 16, 2016, 06:49:49 PM »
0
Good point. I added a note about that in my post.
It isn't about winning the debate. It's about the truth.

TokenMinorityReport

  • FDR Aware
  • **
  • Posts: 89
  • Respect: +142
Re: Actual statistician crushes Molyneux.
« Reply #4 on: February 16, 2016, 07:06:52 PM »
+3
I do want to correct him.  Molyneux was never a "Philosophy major"  ;D

I totally can see him blaming the errors on his "researchers" if ever confronted with this stuff.

This guy got it covered.

AMO 52 minutes ago
 This is outrageously misrepresentative of Molynuex ! He's not a Philosophy major, he took history and then theatre or acting (or something similar) but never philosophy. Not that his absolute lack of training shows, seeing as he has personally solved some of philosophy's most impenitrable conundrums using little more than a set of dumbbells, shouting at people and confirmation bias.
 Reply  · 3
   
"Nor have you built a strong body that you can proudly possess, because such a body is capable of inflicting great violence."--Recently Banned

The Observer

  • FDR Authority
  • ****
  • Posts: 268
  • Respect: +136
Re: Actual statistician crushes Molyneux.
« Reply #5 on: February 16, 2016, 07:28:54 PM »
+1
wow, this is fantastic and hilarious!  Much better than my videos.

skeptic

  • FDR Curious
  • Posts: 3
  • Respect: 0
Re: Actual statistician crushes Molyneux.
« Reply #6 on: November 28, 2016, 09:17:25 PM »
0
Isn't this YouTube character violating his own rules? At ~9:08 into the video the picture in the attached photo appears.  This graph suggests that there is an inverse relationship between poverty and the three types of crime which are included.

Poverty falls as murder, property crime, and violent crime rise.  As poverty then rises (ca. 1998), all forms of crime are reduced.  Does this guy's (unsourced!) graph imply that we must increase poverty in order to decrease crime?  This is an awful thing to suggest!

Re: flipping the axes: the line represented by whichever sort of linear regression was used by Molyneux will indicate the same trend no matter the orientation of the axes.  I can't argue that linear regression was an appropriate treatment, but I can easily and truly claim that the orientation of axes has no impact on the implications of a trend line. 

The referenced video's publisher also does a poor job of explaining why the fact that a quadratic fit reveals two data points with the same "y-value" implies that the data and conclusions arrived at are not meaningful.

What say you all?  Where have I gone wrong?

Thanks!

« Last Edit: November 28, 2016, 10:18:31 PM by skeptic »

Jim Jones

  • FDR Wizard
  • *****
  • Posts: 508
  • Respect: +343
Re: Actual statistician crushes Molyneux.
« Reply #7 on: November 30, 2016, 06:07:45 AM »
+1
Isn't this YouTube character violating his own rules? At ~9:08 into the video the picture in the attached photo appears.  This graph suggests that there is an inverse relationship between poverty and the three types of crime which are included.



The only inverse relationship is the first few years of the 1960s, apart from that, you get a generally flat/positive correlation between poverty and the other indices.

Remember that the YouTuber says something to the effect of "hmm, it's almost as if something happened in the 60s to make this graph misbehave", and you see on the graph that he means the War on Poverty. He doesn't need to prove that is the cause, but he's just pointing out a possible reason. With or without that consideration, poverty isn't really inverse to any index for the bulk of that graph.


Look at the graph again from the start of the War on Poverty (which started in 1964):



Do you still see an inverse relationship?

Also, on the other end, some armchair theorizing on my part for the sharp uptick in poverty without more crime (circa 2008) could be due to the mortgage crisis, wherein those thrown in poverty at that time aren't typically used to a lifestyle of violence despite their situation (this is complete conjecture on my part, just a theory).

Those two theories aside, I think the guy's main point was that there just isn't enough data to prove anything concrete and that Stefan was grasping at straws, trying to make things fit where it'd help him push his agenda.

Does this guy's (unsourced!) graph imply that we must increase poverty in order to decrease crime?  This is an awful thing to suggest!


The source is in the link in the description. I'm guessing he plotted the points himself from the data.

Also, nowhere in the video does he or his graph imply that we increase poverty in order to decrease crime. The graph implies no imperative, it only shows the data.

Re: flipping the axes: the line represented by whichever sort of linear regression was used by Molyneux will indicate the same trend no matter the orientation of the axes.  I can't argue that linear regression was an appropriate treatment, but I can easily and truly claim that the orientation of axes has no impact on the implications of a trend line. 


He wasn't arguing the implications of the trend line based on the axes, but rather how flipping the axes reinforces his point about how Stefan shouldn't have used linear regression since it's misleading. There's a huge number of data points clustered together which paint a conflicting picture, but a very few points on the other end throw the whole thing off. Notice how much of a difference this can make:



The referenced video's publisher also does a poor job of explaining why the fact that a quadratic fit reveals two data points with the same "y-value" implies that the data and conclusions arrived at are not meaningful.


Yes, I don't think he went far enough to explain why. The reason they are not meaningful is because they point to two very divergent points of data and it would be a fool's errand to try and forecast/predict anything based on it.

For example (of my own, not specifically the graph), from 1950 to 2000, you record the level of poverty and crime in the USA. From 1950 to 1975, poverty decreases from 20% to 10%, and crimes increase from 100 to 200. In 1958 the level of poverty is 15% at which point the number of crimes committed is 150 per year.

Then from 1975 to 2000, poverty increases from 10% back to 20% while crimes keep increasing from 200 to 300. In 1983, the poverty level is at 15% and crimes committed per year are 250.

You now have 2 sets of data for the same poverty level despite the fact that it is in the middle of both a decreasing and an increasing trend. The information is not useful for making any forward-looking projections.

This is a very rudimentary MS Paint graph with no time but:



Hope that helps, let me know if I missed anything.


"A system of morality which is based on relative emotional values is a mere illusion, a thoroughly vulgar conception which has nothing sound in it and nothing true." -Socrates

Monster

  • FDR Authority
  • ****
  • Posts: 342
  • Respect: +218
Re: Actual statistician crushes Molyneux.
« Reply #8 on: November 30, 2016, 08:52:35 PM »
0
@Jim Jones

Damn!

skeptic

  • FDR Curious
  • Posts: 3
  • Respect: 0
Re: Actual statistician crushes Molyneux.
« Reply #9 on: November 30, 2016, 10:27:03 PM »
0
Quote
Also, on the other end, some armchair theorizing on my part for the sharp uptick in poverty without more crime (circa 2008) could be due to the mortgage crisis, wherein those thrown in poverty at that time aren't typically used to a lifestyle of violence despite their situation (this is complete conjecture on my part, just a theory).


That's sort of the point, right?  An increase in poverty within a population that isn't very capable of committing crime does not lead to an increase in crime.  The causal link is broken if your conjecture is true.

Quote
Do you still see an inverse relationship?


Nope (unless we use a second order fit which is arbitrary).  I even mapped out the period between 1970 and 2007 myself and found that the rise in crime was larger than the rise in poverty, but the difference isn't much and is probably due to other factors.  I concede that one.

But:



The local maxima and minima seem to show something interesting.  The poverty rate tends to lag behind the crime rate.  There are certainly other factors (it would be interesting to research what was going on in the USA between '83 and '87), but if we are looking at just these two then I'm going to have to side with Stefan I guess.  Re: '83 and '87: There's one instance of a local maximum in crime rate following an inflection point in the poverty rate (the purple arrow), but that is surrounded by two minima on the crime side.

The data I used for "Violent Crime rate" came from www.ucrdatatool.gov and is the combined rate per 100k of murder and non-negligent manslaughter, forcible rape, robbery, and aggravated assault.  The poverty rate is from census.gov and seems to square with the same data everybody else uses.

Quote
He wasn't arguing the implications of the trend line based on the axes, but rather how flipping the axes reinforces his point about how Stefan shouldn't have used linear regression since it's misleading. There's a huge number of data points clustered together which paint a conflicting picture, but a very few points on the other end throw the whole thing off. Notice how much of a difference this can make:


I think that Stefan's entire point in the segment featured is that there is no correlation between gun ownership and the murder rate.  Your statement supports this claim.  Using a linear regression on the plot displayed is worthless, as you've mentioned.  I believe that Stefan's primary motivation may have been to show the R=-0.05 correlation rather than any sort of linear trend.  If, in this segment, Stefan tried to make the argument that there is actually a negative correlation between gun ownership and the murder rate... then he was blowing smoke!



In any case, shifting a few points around in a clearly non-linear data set does nothing to refute my claim that whether we display a set of data in a mathematician/physicist/engineer's preferred style or in an economist's preferred style is irrelevant as far as Stefan's critic's argument goes.  If there is a clear trend then the orientation of the (in)dependent variables isn't important and the linear regression will represent the relationship.  If there is no clear trend then the linear regression is worthless in either case.  The orientation of axes is a matter of style.  I prefer that the independent variable is on the horizontal, but... meh?  It could be reasonably argued that, as mentioned above, Stefan's goal wasn't to illustrate a linear trend, but to show that there is no trend.  Maybe I've missed something.  I just can't imagine a scenario in which flipping the axes matters!

Quote
This is a very rudimentary MS Paint graph with no time but:


By flipping the axes it seems like you've tried to relate this regression analysis argument to an argument about single valued functions.  A mathematical function (in the common sense) cannot map a single input to multiple outputs (at least if it is to be used in analysis [strictly mathematical analysis]), but there is no reason that a statistical correlation cannot display such behavior. 

Hypothetical e.g.: Gun ownership increases gun homicides from zero ownership up to a certain point.  Clearly there will be zero gun deaths if there are zero guns.  Perhaps gun deaths keep increasing until ~80% of the population own guns.  At this point the trend may begin to level off.  If the gun ownership somehow reaches 100% of the population and the majority carry daily, everybody might realize that brandishing a firearm in offense is a bad idea, so the gun crime will fall.

This relationship displays a gun crime maximum at some point between the extremes of the gun ownership domain: zero gun ownership -> zero gun deaths necessarily and 100% gun ownership -> fewer gun deaths than in an environment in which aggressors can reasonably assume that their target is not armed.  This is an extreme gedankenexperiment and I make no claims to truth regarding the specifics, but I don't believe that such situations are inconceivable at all.  That a dependent variable exhibits an extremum which is not at either end of the independent domain is obviously possible.

Don't get me wrong--Stefan Molyneux can be a pompous ass.  I'm just here playing devil's advocate to see what I can come up with because some of his arguments aren't weak.  His politics only appeal to half of the people but, e.g., his idea that violence against children sets those children (and our society thereby) up for failure seems pretty legit.  I don't like seeing folks attempt to blind their opponents with science, and the video posted in this thread is at least as guilty as Molyneux has ever been (maybe more?).

Thanks for taking the time to respond!

Edit to add: for clarification on the MS Paint sketch: it is true that in your example we cannot use poverty rate to predict crime rate.  We can, however, use crime rate to predict poverty rate.
« Last Edit: December 01, 2016, 07:13:11 PM by skeptic »

Jim Jones

  • FDR Wizard
  • *****
  • Posts: 508
  • Respect: +343
Re: Actual statistician crushes Molyneux.
« Reply #10 on: December 13, 2016, 06:02:45 PM »
0
Forgot about this thread. I'll definitely read through this and reply, thanks for the response, Skeptic.
"A system of morality which is based on relative emotional values is a mere illusion, a thoroughly vulgar conception which has nothing sound in it and nothing true." -Socrates