Visualize This: Poverty Rate By Age in America (Jan 14 to Jan 27)

FlowingData Forums » Data Visualization

Visualize This: Poverty Rate By Age in America (Jan 14 to Jan 27)

Started 3 years ago by nathany / 84 posts

« Previous 1…345 Next »

JonPeltier
Member
Posted 3 years ago #

I hate to be a curmudgeon, but...

Like many visualizations, Luca's was very attractive. Unfortunately, like all too many visualizations, it did not readily convey information. DC was very evident at one extreme, but as someone else pointed out, comparing one metropolitan area to whole states is apples-and-oranges. I wonder how NYC, Los Angeles, or Detroit would compare to DC. The rest of the data was lost in a Pollockian blob of paint. I even downloaded the image and viewed at 150%, but most data was hopelessly confounded.

Maybe I'm old-fashioned, but for my money, visualizations should be informative first, and artistic second, and the art should not obscure the information.

I felt that the animated bar chart maps, where each bar had the cross-sectional shape of the state it represented, were also uninformative. I do not care for animations: the data should be present so the viewer can absorb it at his own pace and in the order he chooses. These maps had no guide to what the thickness of each state represented, and many times, tall states obscured their shorter neighbors.

I guess my opinion is not widely shared. Maybe I'm becoming the Dr House of visualizations.
Luca Masud
Member
Posted 3 years ago #

Jon Peltier, as I already said before you are totally right.
The visualizations should be informative first and artistry should never live at the expenses of good information.
But again, I'll stress the fact that in that graph I just wanted to stress D.C.'s exception (hence the title I gave to the map).

It's all a matter of aims: if you want to create a useful tool for comparing data you're going to build a bar chart (since even a line chart is not the appropriate one for this kind of data) but I simply chose not to do that. I wanted to experiment. For sure there are lot of things I can improve, lot of things I could have done better (the blobs are not helpful at all, at least not like this. Simply dots would be much better) but that was just an attempt to do something different. After all the briefing said absolutely nothing about what was the point of interest, what kind of message should be given, what kind of target the graph should be addressed.

Believe me, I do care for integrity, and I'm going to explore the boundaries between graphic design and data integrity in my master thesis. I just still have a lot to learn. =)
ben
Member
Posted 3 years ago #

@nimao

Really attractive presentation of the data. I would love to see this visualization as a large format print poster -- it is pretty and visceral. The dolls are a really haunting image. It also could be re-fashioned into a cool video.

One suggestion would be that you write the actual percentages down on the polaroids, and maybe one other state fact to reward a person that wants to get up close to each photo. e.g. "Mississippi has the highest average poverty of any state".

Nice job!
nathany
FlowingData Editor
Posted 3 years ago #

@JonPeltier Definitely from an analytical standpoint, Luca's does have flaws e.g. the blobby-looking distributions. For best analytical entry, I'd go with Hadley's or maybe yours which is essentially Hadley's with lines. I haven't decided if I like the bulleted lines. I know you're a big fan of them.

What I did like about Luca's though is that he did a little bit of sifting, found something he thought was interesting, and then rolled with that. I also appreciate the incorporation of GDP.
marclevy
Member
Posted 3 years ago #

Sorry, didn't see this thread until the window was closed. But here's a map using higher-resolution census data that may be of interest.

This and related maps are available here: http://sedac.ciesin.columbia.edu/usgrid/maps.jsp

Marc Levy

[attachment=345,83]
cloister
Member
Posted 3 years ago #

@marklevy: Wow. That first map screams a wholly different fact. "The Navajo reservation is a pit of poverty." But the the second one shows that while the percentage rate in that area is very high, hardly anyone lives there.

Could you do a map that multiplies the two values and maps them onto some useful scale?
nimao
Member
Posted 3 years ago #

@Ben

First of all, thank you Ben
However the map is designed to be implemented on a large panel and smile :( next to the name of the state in the white space under the polaroid indicate the total percentage of the state (perhaps it would be better to use numbers but I wanted avoid as much as possible the use of the numbers).
However they are all things that I simply forgot to write cause I finished the map very late... :)
nathany
FlowingData Editor
Posted 3 years ago #

@nimao - in any case, very nice (and creative) work :)
hadley
Member
Posted 3 years ago #

I'm also surprised by the number of very attractive but not terribly informative visualisations. Many of them present a strong moral message rather than letting the data shine through. This probably reflect a fundamental difference in how authors imagine their work being used. I picture my graph being used in a meeting of sober government bureaucrats deciding where to spend their money to have the most effect. Others obviously imagine their work being used to rally the public to fight the evils of poverty.

I also wonder at the absence of data exploration - is it not interesting the child, adult and elderly poverty rates are all roughly the same proportion of the overall poverty rate across states? My feeling is that most people spent a lot of time on a single representation of the data, rather than exploring many possibilities.
nathany
FlowingData Editor
Posted 3 years ago #

@hadley were you really that surprised? FD readers are pretty diverse in that sense, so I pretty much expected the mix, which I really like. It's that mix of emotion and logic that I think is the most interesting.
dsimmons
Member
Posted 3 years ago #

@all I had some ideas about what to do for this. I'm glad I didn't spend the time, they would have been total 'yawners' in comparison. Well done, all.

@Nathan,

please, please don't have different threads for different submissions. Yes, it gets a bit cluttered and lengthy and that's anti-thetical, but there are many more of us that scroll through the thread admiring the great work. The many more of us would have to go back to every thread to check them.

The habit of using the @sign is a great solution
cloister
Member
Posted 3 years ago #

@hadley: "Many of them present a strong moral message rather than letting the data shine through."

This is starting to get pretty philosophical, and I think that it would be nice to see Nathany write up a blog post about it and ask for comment from the whole community in a separate thread, but:

I have to ask, is there really any difference between presenting a strong message (I edit out the word "moral" because the message doesn't necessarily have to be a moral one) and "letting the data shine through"? I mean, really. What's the difference?

Case in point. One of Tufte's classic examples is the O-ring burn-through data presented to NASA decision makers before the Challenger accident in 1986. Engineers knew that O-rings had a problem with eroding due to hot gasses within the solid rocket boosters, and that this problem was worse when the weather was cold.

Not being avid readers of FlowingData, when they presented the data, they organized it by time--launch date--rather than by ambient temperature at the time of launch.

Since launch temperatures are basically un-correlated with the calendar date, the pattern of "more burn-throughs when the weather is cold" was just not visible. If memory serves, Tufte reproduces the visuals from the presentation in his first book to make that point, and he's right. The decisions makers couldn't see it, and ordered the launch to proceed.

Re-organize the data with temperature as the x-axis variable, and the trend is as clear as the cold winter day that killed those seven astronauts and grounded the shuttle fleet for the next several years.

The engineer's message--indeed, the message that the data held--was obscured due to poor visualization. The data didn't, in your words, "shine through". And neither did the message.

To me, that's what "letting the data shine through" means: making the message within the data as obvious as possible. It means turning raw data into knowledge that can be acted upon.

Otherwise, what's the point? If all you want is to show the data, just print out the raw spreadsheet. Data galore, but completely useless.
nathany
FlowingData Editor
Posted 3 years ago #

Hmm, I'm actually more with Hadley on this one. Having worked with statisticians (uh, I am one) and designers, there's a pretty big difference between letting the data shine through and highlighting something in the data.

The NASA example supports what Hadley said about EDA. The engineers, the initial analysts, or whatever looked at the data from one angle, but if they had explored a bit, then they would have gotten some different results.
Luca Masud
Member
Posted 3 years ago #

Yeah, but it simply depends on what you want to say and communicate. There may be a lot of data in a phenomenon but usually visualizing and talking about all of them can be completely useless. There's design before visualization: you have to make decisions about what you want to highlight with your graph and then starting graphing in the proper way to deliver your message.
That's what I think Cloister wanted to point out.

Obviously this doesn't mean that you have to do poor graphs. You can perfectly stay "true" to every single rule of good graphing but still decide to convey a certain message by not displaying some of the data that is not useful to solve your problem.

It's all about the starting question.
JonPeltier
Member
Posted 3 years ago #

@Nathan:

"[1] What I did like about Luca's though is that he did a little bit of sifting, found something he thought was interesting, and then rolled with that. [2] I also appreciate the incorporation of GDP."

2. GDP was a nice covariable to incorporate. I didn't have time to think of anything to could look up to add to the analysis.

1. Again, what I find noteworthy is that a geographically small metropolitan area did so much differently than the geographically much larger states. The second map posted by Marc Levy brings up this point. Based on number of people in poverty per square mile, I see a large number of dots which are larger than DC: NYC, LA, Chicago, Detroit, Phoenix, Portland OR, Houston, DFW, to pick off a few. This measure may not be comparable with the percentages in the file you supplied, but it shows that the DC data is out of place with the states data. The fact that it was such a blatant outlier in Luca's graphic should have also demonstrated this.

And now I'll shut up!
hadley
Member
Posted 3 years ago #

@nathany: yup, definitely surprised - the majority of submissions look like they have been created by artists not scientists or engineers. I'm not saying that this a bad thing, but it worries me that there appear to be few more quantitative people using visual tools and practising their visualisation skills (maybe they are all too busy doing more important things)

@cloister: it's easier to speak in specifics than generalities, and in this case I think many people came to the challenge with a preconceived notion of what the visualisation should show: that poverty is bad and we should do something about it. That may be appropriate in this case, but in most cases it is not: you need to understand the data before trying to summarise them in a simple fashion. You need to be clear headed and dispassionate in an analysis, and not immediately jump to an emotive presentation.

@luca: of course you have to make decisions about what to highlight in a single graphic. However you can only make a good choice if you have considered many possible options. If you present your first thought you may miss many important features. The design must come after the analysis - you will only find out the message after trying many possible graphics (otherwise you are just producing propaganda).
nathany
FlowingData Editor
Posted 3 years ago #

RE: technical participation, yeah, that worries me too. I'm just going off feeling, but it feels like that the technical crowd really likes to go off defaults on whatever their program gives them.
cloister
Member
Posted 3 years ago #

@nathany:

Fair enough, but then I again have to wonder what you guys mean by "letting the data shine through."

Can you explain what that means in such a way that is distinct from "presenting data so as to convey a particular fact or insight"?

Not being a statistician, I'm not seeing it. Which is not to imply that it isn't there; I'm a communications guy myself, so for me there's no reason to have data unless you can learn something from it, and communicate that with others.

I'm not trained to look at data like a statistician, so I'm genuinely curious as to what the alternate viewpoint is.
Luca Masud
Member
Posted 3 years ago #

@hadley: I did try many different graphics. But, after all, we all didn't have much time to work more on data and since I'm not an analyst maybe I'm much slower than you in seeing data correlations. For sure there were other things one may have considered but when I incorporated the GDP I was thinking, someway, that I was playing "unfair" by putting more data than the one given (which I found - no offense, Nathany - a little uninspiring: no vertical insight at all). But I used it anyway because it seemed to me a good intuition.

What says Cloister is truth: it seems to me that staticians, or people like Jon Peltier (from whom I have a lot the learn) sometimes forget that not everybody of us are able to read data and the many nuances of graphs. I'm trying to learn but I cannot expect that every people in the world is going to.

That's why data integrity is indeed important but, still, you always have to remember who are you talking to (the target).

For example, let's take the case in which the graph is aimed to explain something to generic people and it's going to be published on a magazine or newspaper.
In this case it has absolutely no sense designing a perfect graph that is great in understanding the system when read by staticians. Much better is communicating what the staticians saw in that graph, taking the hand of the reader and guiding him through all the data, telling him what to look at. Obviously you can do it in a bad or a good way, and one must try to avoid all kind of data distortions and gross mistakes (like trying to compare thousands of pie charts! =P ).
But still, in that case, data really comes after the message you want to convey: what's the point of a perfect but useless graph?. And I rather prefer showing less data and going straight to the point (I'm not trying to defend my graph here, it is indeed full of mistakes).

Of course the opposite is true when the target works as a tool and it's aimed to staticians, analyst, etc. They do know how to read graphs. They do know how to see correlations. That's why they surely don't need strong guidelines as would do a generic magazine reader.

And sorry if my English is not that good.
JonPeltier
Member
Posted 3 years ago #

@Luca -

Don't apologize for your English. You communicate your ideas clearly.

You're absolutely correct, that the display has to be matched to the audience. But I think we don't give "generic people" credit. If the visual is clear and the textual description is adequate, people will understand more than publishers generally seem to expect. A little color is good to get their attention, but one must take care not to let the design obscure or distort the information. I discussed that on my blog in Bad Bar Chart Practices, or Send in the Clowns.