Personally, I've found that any time someone declares that a given visualization "speaks for itself" what they mean is that they want you to not ask questions about it. Occasionally this is justified, but more often it's a stunt of some sort.
PRs welcome! The source code is here: https://github.com/statwonk/paygap
I appreciate the gesture. Unfortunately, statistics is not my strong suit and R is not a language I am familiar with.
That said, did you consider the social and media context into which your work is being injected before you did so?
"Here's some misinformation for everyone! Fix it for me or keep your comments to yourself."
I am noticing that the pulled list of occupations and the generated output differed in size. Why is that?
For example I would like to see the data for "Automotive Body and Related Repairers".
The difference are non-published values. Really wish we had raw surveys, but I do understand the ethical challenges BLS deals with.
I have a feeling this graph is going to be reposted across various social medias with less context and incite more people spreading misinformation.
Absolutely agreed. It's irresponsible to present incredibly nuanced data at such a shallow face value without making any attempt to show the full story, especially on such an incendiary and divisive topic. How is this helping clear up disinformation and progress gender wage gap dialogue?
That most occupations pay women less than men is a fact, the graph is just a reformulation of this table: http://www.bls.gov/cps/cpsaat39.htm
Facts do not absolve a data visualization from being potentially misleading.
I'm sorry you feel that way. I sincerely find it very informative. Before I got first glimpse of it, I thought there would be at least 10-15 professions where women edged out men.
This isn't just a feeling. The existence of phenomenon such as Simpson's Paradox mean that this is something that any data analysis must be very careful about. Presentations of aggregate data are particularly vulnerable.
I'm very well aware of Simpson's paradox, it was why I wanted to look at the data broken down by occupation in the first place. :)
That's never the case with data visualization. Explanation helps remove ambiguity, and the lack of explanation behind the data source is why there is an excessive amount of criticism in the HN comments.
Regardless, the graph does not certainly speak for itself. There's no obvious correlation between which professions have a better gender pay split than others. If you're trying to make the point that most professions pay women less than men, then you need to justify that. (And provide evidence that the chart represents "most professions.")