Show HN: Sex, politics, and pay gaps in the U.S | Hacker News Reader

Show HN: Sex, politics, and pay gaps in the U.S

visualmagnetic.com

17 pointsakarve14 comments

Nice map/presentation, but I fail to see what relationship these metrics are suppose to have with one another. There' doesn't appear to be a strong correlation between them and it comes off as rather sophomoric.

author here. a more rigorous analysis would include a co-variance matrix and a best fit line to quantify the relationships between the variables.

on the surface, there doesn't seem to be much of a correlation. not all exploration reaches a definitive conclusion or even a correlation. the alternative would be for me to shelve the visualization on the grounds of "no conclusion". and that would be the file drawer effect in action :) http://en.wikipedia.org/wiki/Publication_bias#File_drawer_ef...

OK, but what made you pick those three variables to start with? I can see political leaning vs pay gap, because equal pay legislation/administration is a somewhat partisan issue. But the length of sexual intercourse measure doesn't seem to tell us much about anything. Why not measure annual consumption of corn dogs, or the going price of onions?

I'm open to there being a good reason you chose to include it, but you didn't offer on on the page so I'm just confused; it reads as a sort of really abstract titillation, even though it could be interesting to learn what it does correlate with.

i saw nerve.com's data on intercourse duration and started to wonder which variables might cause or correlate with the same. of greatest interest to me at the time: how does in/equality affect our sex lives? income is one measure of equality. politics is a natural third dimension of general interest.

it is a somewhat ad hoc trio of variables, sure, but there's a singular underlying question: which popular psychographic dimensions affect our sex lives? given finite time, i picked these 3 and ran with it :)

Thanks for taking the time to explain. Although I was critical of that aspect I thought the design itself was clean and effective, and you should keep exploring this line of activity.

There's no need to shelve the entire visualization simply because the metrics are not relevant (I hope). It looks like it's built with D3, so it shouldn't be too much effort to swap out the data with something that actually does have value.

By the way that's seriously misrepresenting the file drawer effect. Your effort has been spent building the visualization, not doing independent research. If you spend some time on statistical analysis of something that can be reasonably expected to be correlated and find that it's not, please do publish that information. But that's simply not the case here.

Agreed. The visualization doesn't help to identify whatever correlations there may be.

Data visualization should be informative and intuitive, at its basest. This is neither, especially with the distracting, unnecessary animations.

Data visualization can be well designed, yes, but it's not necessary, and it should definitely not come at expense to ease of interpretation.

hi. how would you suggest making it more informative & intuitive, aside from removing the animations?

What is your purpose in graphing these three metrics at the state level? What are you trying to communicate?

In addition to the criticisms raised by others (the most important of which is significance!), I'd add that the dots are a poor way to demonstrate the pay gap. People are notoriously bad at estimating area, particularly when the variance is so small. The only reason pie charts are acceptable is that you double encode the metric in area and central angle.

Here you are trying to demonstrate the ratio between two values: the average pay of women versus men. However, the dimension encoding this metric is overall area, so the fact that this is a ratio is obscured. Perhaps worse, the areas are misleading: WY pays its women 37% less than men, yet the dot is barely visible, far smaller than 63% of the area of the circle (which would be the intuitive explanation); likewise, it looks (on the scale to the right) that even 100% pay equality is represented by an incomplete circle.

Also, it looks like you're plotting DC but it's basically impossible to click on to see the data; for that matter RI, CT and MA are pretty hard to deal with.

It seems like you're using a force-based layout for the charts, which I take to mean that the states aren't accurately placed. Please don't move data points around to suit aesthetic concerns.

Okay, enough negativity, some positives. I quite like the scales to the right. Some may complain about the animation, but I think that's valuable object constancy in action. It makes it quite intuitive what's happening when rolling over the various states. I also like the use of left/right to represent the political lean.

However: it isn't at all clear where the ends of the scales are and why they were chosen. The arrowhead makes estimating the length of the political lean bars difficult. The fact that they are offset by the width of the circle makes comparison around the middle impossible, and comparison elsewhere hard.

Is this data geographically related? If not, a choropleth is probably not the right solution. The fact that you're charting state-based data is not enough to justify a geographical representation.

EDIT: the most important aspect of a data visualization is the message. Don't just try to make something pretty: think long and hard about what you are trying to convey. And add text! An explanation is required to make an interesting visualization comprehensible. If a visualization is completely understandable without an iota of text, the message is probably entirely obvious without the visualization.

The "pay gap" statistic does not control for occupation, age, or education. What's the purpose of such an apples-to-oranges comparison?

If anyone is interested in a similar basic crossfilters written in d3, you can check out dc.js - https://github.com/dc-js/dc.js/wiki (edit: i have no relation to the project, but it's pretty neat)

Here is a list of community contributed examples: https://github.com/dc-js/dc.js/wiki#examples-contributed-by-... including some geographic visualizations.

You could also simply work with Crossfilter [0] and D3 [1] directly. They're both great libraries, and learning their APIs independently seems more valuable.

[0]: http://square.github.io/crossfilter/

[1]: http://d3js.org/

Reply on news.ycombinator.com