Hermit wrote:
I provided links to the raw data the chart is based on. Feel free to show me where the problem lies, but if you decide to do that, cite data and the process you used to prove my chart wrong. Hand-waving - be it yours or anyone else's - has never convinced me.

You say you used those data to make the chart. I don't know which data you used, such as which specific table columns you (think you) plotted. Why don't you just bag it and go home if you don't want to tell us, or you didn't keep any records? You don't understand how this stuff is done, and you clearly do not respect those who are telling you to publish your fucking methodology. If you want to become somebody your intended audience respects, and you really think this argument is worthwhile, you'll publish your fucking methodology, which is something that allow someone else to replicate your alleged experiment.

Cito di Pense wrote:You say you used those data to make the chart. I don't know which data you used, such as which specific table columns you (think you) plotted. Why don't you just bag it and go home if you don't want to tell us, or you didn't keep any records? You don't understand how this stuff is done, and you clearly do not respect those who are telling you to publish your fucking methodology. If you want to become somebody your intended audience respects, and you really think this argument is worthwhile, you'll publish your fucking methodology.

The links to the data are published in the post where I originally placed the chart, as is a description of the "fucking methodology" I used. You actually replied to that post, so I'm sure you'll have no problem finding it before shooting your mouth off once again, pretending you know what you're talking about. If you can't be bothered chasing it up, I expect to see more of your hand-waving rubbish.

Hermit wrote:

Cito di Pense wrote:You say you used those data to make the chart. I don't know which data you used, such as which specific table columns you (think you) plotted. Why don't you just bag it and go home if you don't want to tell us, or you didn't keep any records? You don't understand how this stuff is done, and you clearly do not respect those who are telling you to publish your fucking methodology. If you want to become somebody your intended audience respects, and you really think this argument is worthwhile, you'll publish your fucking methodology.

The links to the data are published in the post where I originally placed the chart, as is a description of the "fucking methodology" I used. You actually replied to that post, so I'm sure you'll have no problem finding it before shooting your mouth off once again, pretending you know what you're talking about. If you can't be bothered chasing it up, I expect to see more of your hand-waving rubbish.

What you wrote doesn't permit anyone to replicate your alleged experiment and check to see if you made mistakes in your analysis. You're asking your audience to guess which specific data you plotted among what you linked. You claim you're old enough to know better than to pull a dodge like this one. Commit yourself to identifying the specific data you plotted. Can't do that? You probably have no record of what you did, unless you still have your alleged spreadsheet. If you have that, you can make some progress.

Cito di Pense wrote:

Hermit wrote:

Cito di Pense wrote:You say you used those data to make the chart. I don't know which data you used, such as which specific table columns you (think you) plotted. Why don't you just bag it and go home if you don't want to tell us, or you didn't keep any records? You don't understand how this stuff is done, and you clearly do not respect those who are telling you to publish your fucking methodology. If you want to become somebody your intended audience respects, and you really think this argument is worthwhile, you'll publish your fucking methodology.

The links to the data are published in the post where I originally placed the chart, as is a description of the "fucking methodology" I used. You actually replied to that post, so I'm sure you'll have no problem finding it before shooting your mouth off once again, pretending you know what you're talking about. If you can't be bothered chasing it up, I expect to see more of your hand-waving rubbish.

What you wrote doesn't permit anyone to replicate your alleged experiment and check to see if you made mistakes in your analysis. You claim you're old enough to know better.

You asked for which specific table columns I plotted, and what methodology I used. I provided them in my original post on the topic, and I laid out the methodology by which I came to produce the chart in that same post. There is certainly enough information for anyone to replicate and/or repudiate what I posted, but I don't much care that you are incapable of doing either.

I must admit, I'm not quite getting the graph either. It appears that when approximately 5% of households own guns there are either 2.3 murders per 100,000, or about 7/100,000. At 40% ownership we get everything from just over 1/100,000 to nearly 10/100,000. If we take all guns out of circulation there are no murders, period. Nor are there at ownership rates in excess of about 60%.

If this data is correct it appears that gun ownership has absolutely nothing to do with murder rates.

Hermit wrote:
You asked for which specific table columns I plotted, and what methodology I used. I provided them in my original post on the topic, and I laid out the methodology by which I came to produce the chart in that same post. There is certainly enough information for anyone to replicate and/or repudiate what I posted, but I don't much care that you are incapable of doing either.

What you're doing now is what is commonly known as obfuscating your 'data'. You might think you did some 'research', but you're not getting a lot of takers. Your purpose is clear in not revealing which data vectors you (think you) actually plotted, of which there should be only two, to make a 2D correlation. What a joke on your audience. Well done, you.

laklak wrote:I must admit, I'm not quite getting the graph either. It appears that when approximately 5% of households own guns there are either 2.3 murders per 100,000, or about 7/100,000. At 40% ownership we get everything from just over 1/100,000 to nearly 10/100,000. If we take all guns out of circulation there are no murders, period. Nor are there at ownership rates in excess of about 60%.

If this data is correct it appears that gun ownership has absolutely nothing to do with murder rates.

There aren't any examples of locales with greater than 60% gun ownership. It sure would be interesting if there were. The NRA is seeing what it can do about ameliorating the situation. The graph is "un-gettable" since there isn't a regular scale for homicide rate. That figures, though, because homicide is an irregular event. Highly irregular.

laklak wrote:I must admit, I'm not quite getting the graph either. It appears that when approximately 5% of households own guns there are either 2.3 murders per 100,000, or about 7/100,000. At 40% ownership we get everything from just over 1/100,000 to nearly 10/100,000. If we take all guns out of circulation there are no murders, period. Nor are there at ownership rates in excess of about 60%.

If this data is correct it appears that gun ownership has absolutely nothing to do with murder rates.

There is no constant ratio between gun ownership and homicide rates from state to state. There are just too many complicating factors involved for that, such as the urbanisation rate, average income levels, and whatnot others of one state compared to another. Unsurprisingly you may find two states with very similar levels of gun ownership sporting quite different homicide rates. For example, Michigan (28.8 & 5.8), Vermont (28.8 & 1.6).

Altogether, a trend is not readily discernible by eye. That makes it easy to "prove" anything you want to via some judicious cherry picking. If you use the data of all states, however, cherry picking is no longer possible, and that is what I did. The y-axis measures the percentage of households (which should explain to you why the scale doesn't go much beyond 60%) of all states, and the x-axis measures murders/100,000 of the population of all states. Since we're dealing with percentages and per population, the differences in population between states is irrelevant. I then let the spreadsheet's algorithm work out a trend. It came up with a slight upwards gradient - despite the sawtooth pattern of gun ownership the trend is an increase of murder rates accompanying an increasing percentage in gun ownership per household.

I can assure you that I did not massage the data in any way. All states are there and the two relevant columns (gun ownership per household and murder rate per 100,000) were a pure copy-paste job. As for the trend line itself, that's just a matter of clicking a button. You can replicate what I have done yourself if you could be bothered. I provided links to the data and described the way I entered them in the spreadsheet. You could even resort the columns, grading by gun ownership instead of murder rates on the x-axis. In that case the same trend could be seen, just differently. Instead of gun ownership, the sawtoothed line would represent murder rates, but the trend would still be a gradient indicating the same relationship.

Had that line turned out to be horizontal, I would have argued that there is no correlation between gun ownership rates and murder rates. Had that line turned out to be dropping, I would have argued that John Lott was right: More guns, less crime.

despite the sawtooth pattern of gun ownership

The sawtooth pattern has no place being there from what I can see. The chart should show individual points without lines drawn from one point to the other, because there is no natural progression from one point to the other, each is an individual case with different circumstances. A trend may mean something overall, but the line is a pointless distraction.

zulumoose wrote:

despite the sawtooth pattern of gun ownership

The sawtooth pattern has no place being there from what I can see. The chart should show individual points without lines drawn from one point to the other, because there is no natural progression from one point to the other, each is an individual case with different circumstances. A trend may mean something overall, but the line is a pointless distraction.

Yes. The sawtooth line makes it look as though there is some sort of connection between those particular data points. It does nothing more than highlight the fact that the correlation between gun ownership and murder rates is by no means straightforward. A scattergram would have been appropriate even if that highlight were lost.

The following figure certainly appears to show the data you plotted:

Now look at it with a linear fit added:

You can bullshit people only so long, Hermit. You threw out the data for Washington DC because it pulls down the regression line, and for no other discernible reason. You should always publish your data massage technique when you give your results. Please note that my plots include a properly-labeled horizontal axis, and includes one data point above an x-value of 10 and one data point with an x-value near 0, which you omitted from your "graph".

All the data are normalized to murders per 100000. You can't toss out a data point just because it pulls down your regression line. Anyway, who said plotting the data state-by-state is a meaningful way to correlate? It's abundantly expected that this sort of wank-fest will have no discernible correlation. This is because the number of gun murders and percentage of gun ownership in one state are not related to the number of gun murders and percentage of gun ownership in another state. That's like saying one state is pretty much like another, and if you look at the US from a good enough distance, like from the moon, it sure might seem that way.

Cito di Pense wrote:You can bullshit people only so long, Hermit. You threw out the data for Washington DC because it pulls down the regression line, and for no other discernible reason.

Bullshit? I included all states. Since when is Washington DC a state? Looks more like a city to me anyway even if it were a state. In fact, Washington DC is not even as a city. It's a fraction of the Washington metropolitan area in terms of population and area. It is also smaller than 19 US cities in terms of population and smaller than 47 cities in terms of area.

Kudos for finally looking at clicking on the links for the data I used that generated the chart, both of which appear in my very first post in this thread.

Now, replicating your mistake i.e. including Washington DC in the calculations, I get this:



Not a huge difference between that and the original one that excluded DC, don't you think?

As long as you throw out the Washington DC data, and nothing else, you'll get the sort of regression line you're getting, and you won't get it any other way, unless you make up something appropriate for the missing data point for one state. So you are lying by saying your upper chart includes it. Now, your only problem is to justify why you've taken it out. It looks to me as if you've taken it out so you'll get the regression you want. Just make a table, listing the states and the relevant data you plotted, and we can compare it with the data in the link you gave. At this point, you're doing nothing but trolling someone who's serious about analysis in a way that you are apparently not.

I finally got around to asking you how it was meaningful to aggregate data by states (however you define 'states', Washington DC has a larger population than Wyoming, even though the former has no representation in the US Congress). You don't want to address that question, do you? If you think about it, though, you'll realize that Washington DC and Wyoming are entirely different sorts of societies, something a sociologist should be able to consider.

The data you chose to plot is not granular enough. I don't care whether you think you know what you're doing, or not. You don't. Oh, and you're still drawing the jaggedy lines, just as if you don't know what the fuck you're doing.

Hermit wrote:Yes. The sawtooth line makes it look as though there is some sort of connection between those particular data points. It does nothing more than highlight the fact that the correlation between gun ownership and murder rates is by no means straightforward. A scattergram would have been appropriate even if that highlight were lost.

That, Hermit? You're still trying to cover up for the fact that you didn't start with a scatter plot and document your methdology. I'm not surprised you didn't; your methodology is in the ether somewhere.

Cito di Pense wrote:As long as you throw out the Washington DC data, you'll get the sort of regression line you're getting. Now, your only problem is to justify why you've taken it out.

I just told you why I omitted Washington DC. Including DC is a nonsense for the reasons I listed earlier: Washington DC is not a state. It's not even as a city. It's a fraction of the Washington metropolitan area in terms of population and area. It is also smaller than 19 US cities in terms of population and smaller than 47 cities in terms of area.

Additionally, as I also pointed out, even including the DC data point makes little difference.

Hermit wrote:

Cito di Pense wrote:As long as you throw out the Washington DC data, you'll get the sort of regression line you're getting. Now, your only problem is to justify why you've taken it out.

I just told you why I omitted Washington DC. Including DC is a nonsense for the reasons I listed earlier: Washington DC is not a state. It's not even as a city. It's a fraction of the Washington metropolitan area in terms of population and area. It is also smaller than 19 US cities in terms of population and smaller than 47 cities in terms of area.

Additionally, as I also pointed out, even including the DC data point makes little difference.

You are lying about that. Show us the table of data you're using to assert that, or shut up.

Are you going to give an argument for why you think breaking the data out by states (however you define them) gives a meaningful correlation? No, I guess you're not going to do that, are you?

I'm really wondering why others participating in this thread, like Olivier, those who are more used to reviewing survey research, and the often-ridiculous relationships 'researchers' derive from it, didn't catch that.

Cito di Pense wrote:You are lying about that.

calling me a liar? Nice one.

Cito di Pense wrote:Show us the table of data you're using to assert that

Huh? You want me to show you the table of data I used after I provided the links to them and after you accessed them?

Cito di Pense wrote:

Hermit wrote:

Cito di Pense wrote:As long as you throw out the Washington DC data, you'll get the sort of regression line you're getting. Now, your only problem is to justify why you've taken it out.

I just told you why I omitted Washington DC. Including DC is a nonsense for the reasons I listed earlier: Washington DC is not a state. It's not even as a city. It's a fraction of the Washington metropolitan area in terms of population and area. It is also smaller than 19 US cities in terms of population and smaller than 47 cities in terms of area.

Additionally, as I also pointed out, even including the DC data point makes little difference.

You are lying about that. Show us the table of data you're using to assert that, or shut up.

Are you going to give an argument for why you think breaking the data out by states (however you define them) gives a meaningful correlation? No, I guess you're not going to do that, are you?

I'm really wondering why others participating in this thread, like Olivier, those who are more used to reviewing survey research, and the often-ridiculous relationships 'researchers' derive from it, didn't catch that.

Because I was in the middle of driving my son 700km so he can spend the upcoming weekend competing in our state's Under 15 field hockey championships, which seems, on catching up on this thread, to have been a superior use of my time.

Olivier, what are you doing having a life?

Hermit wrote:Huh? You want me to show you the table of data I used after I provided the links to them and after you accessed them?

Stick to stuff you know something about:

Hermit wrote:I don't know anything about Blur, Oasis, The Cure, The Prevention or whoever. I only know about Led Zeppelin, The Nina Hagen Band, The Mothers of Invention, The Bluesbreakers and some others nobody talks about any more, so I'm not sure about how best to fan the embers. Please help.