Binge Alcohol Drinking and Suicide Deaths by State

This is an incredible graph. 94% correlated and look how closely the lines track. It is an amazing insight into the data.

Actually, I'd expect there to be a solid correlation. This is absolute data and so tracks population. Note Alaska being very low on the chart but leading the nation in suicides in the rate data.

This graph doesn't show what it claims to show. It should be a scatterplot of suicide deaths (on y) vs binging (on x) with units for both stated clearly. States should be identifiable as individual points in the graph. This form is an example of poor display design

I agree with Dataman. This should be a scatterplot.

you can make it scatter plot yourself by clicking on "more options" above. that is one way to see this data, a line chart is another, a map is another and so on...
http://www.swivel.com/graph...

First off, as to absolute vs relative, if you swap it over, the curve stays approximately the same. I agree that the graph may be a bit deceptive, but I had a hard time finding an appropriate representation for so many states. I believe that some of the problem may be due to data formatting; I was trying to work with a dataset that I made (Suicides) and incorporating a dataset I found on here (Drug Use - Binge Drinking), without reformatting and double-posting datasets, or processing things by hand and converting them to different units. I actually made this graph whilst in the process of working out the various options for graphing that are available. I don't think I was expecting it to be featured. :) Anyway, to move on, I wasn't too pleased with the scatterplot results, not because I was seeking a positive or negative correlation at all (I wasn't), but because it looked a bit junky and not aesthetically all that readable (again, the 51 points per set). All of that aside, if you aren't pleased with the graph(s) and choose to look at the raw data, which was how I got to making the graph(s) in the first place, there *is* a very definite correlation; I made sure of that by clicking through the graph and data options before I featured it personally (I'd be open to hearing if I am wrong, of course). While some of that may indeed be product of comparative population density, on an incidence percentage basis things do line up in a rather disturbing way (and it's doubly disturbing to live in some of the places that aren't ranked high for either trend when they very rightfully seemingly should be; Wyoming winters, for example, aren't pleasant, in my experience). I honestly am not sure there was a pretty way to present this, but pretty was the afterthought for me. I'd love to hear any ideas. Thanks for the comments! :)

unk_variable, I think the reference to absolute versus relative means that the data needs to be based on population counts - so you really can't do anything with the graph that allows that comparison. Each point needs to be per capita. Per capita numbers could show a different relationship - although I would still imagine some correlation to exist.

This was actually what became of the graph I created last month for per capita numbers, if anybody is interested (I fixed it up a bit to post here, but the graph figures are the same): http://www.swivel.com/graph...

As far as I can tell, unk_var, those aren't per capita numbers. It's looks like they establish a correlation between population and suicide rate. This whole thing seems a bit clunky. We would need to set up a graph relating suicides per capita to the incidence of binge drinking. And even that would only go to prove a point most of us suspect to be true. It might be interesting to look at the states that break that pattern, though.

good summary!

Still frustrated with this analysis. The data needs to plot the relative death rate from suicide vs binge drinking to see if there is a correlation. Let me see if I can post something that makes sense.

The exercise, to me, was to see if it could be done with existing data sets. It'd be relatively elementary to do if you don't mind doing math and uploading one or two new datasets. This is where things will get tricky, not just with regards to this graph, but probably many more. What will be even trickier to discern is not really the apples & oranges, but, say, to stretch the metaphor to the breaking point, oranges and tangerines, or even tangerines and mandarins... Most people aren't really willing (and many not able) to search for or create the perfectly corresponding data sets when they're apple and oranging. While I love the raw flexibility of this site, I do foresee it also causing some technical challenges for swivel to overcome. :)

unk -- I combined your data with a data on state population and death rates and there is absolutely no correlation between the relative death rates from suicide and the fraction of the population who admit to binge drinking.

http://www.swivel.com/graph...

Anyway, metaphors aside, the original chart is very misleading and guides the user to an obviously wrong conclusion. We as power users of data and stat analysis need to be very careful about guiding those around us to make valid conclusions with good presentation of graphical information.