Misleading charts of 2015, fixed | Hacker News Reader

Misleading charts of 2015, fixed

qz.com

79 pointsmikek21 comments

I would argue the last "good" chart example, a chart of mass-shootings per president, is misleading as it compares four two-term presidents (Obama, W. Bush, Clinton, and Reagan) to a single-term president (H.W. Bush).

Deleted comment as I was talking about the wrong graph....

The first one was indeed a bad chart, the others simply lying with data. That might as well have been wrong in written text...

There is lies and there is statistics :). For more reading on the subject, I recommend Beautiful Evidence by E. Tufte. He's a classic when it comes to infoviz.

So this complains about not including the 0 on the y-axis but then complains about it in the national review global temp chart.

If you have an agenda it will show through, regardless of what you do. Both graphs accurately portray the data, but most of the public doesn't care to interpret graphs objectively.

You'll notice that in his 'best version' of the education chart, the y axis also does not start at zero. More precisely, the point he made was that _column_ charts should always start at 0. He makes no blanket statements about line graphs to show trending.

Also technically any graph that isn't numerically wrong is a problem of the public not taking care.

Technically according to what rule?

Bar charts use length to show amount while line charts use slope to show change. Therefore, it is misleading to cut off the y-axis for a bar chart, but not for a line chart so long as slopes are transformed proportionally. For line charts, it is generally best to constrain the y-axis such that change (keeping in mind the precision of the data) can be discerned. Of course, if zero happens to be meaningful, it could be illuminating to include and annotate it, just as you would any other relevant point of reference. However, in the case of the temperature chart, zero degrees fahrenheit is arbitrary. Ask yourself, why not zero degrees kelvin?

Agenda will show through. Hopefully, in data viz, your agenda is scientific/statistical accuracy.

Should follow up on the lack of 2008 PP data. Might show up as an abortion peak...possibly why it has apparently been witheld.

Uh..downmods?

You're saying a missing report, probably the result of miss filing the data, is a malicious attempt to hide some sort of abortion spike in 2008.

If you find sources you might get upvotes, but given the available information it's probably simply missing.

Interesting perspective. What harm do you think occurs by 'hiding' such information? (You'll also note that I did not use the word malicious, nor imply its meaning.)

The source of the data would necessarily need to be Planned Parenthood, yet the data for only that particular year is missing.

Anyway, I'm not here for the upvotes; only facts and analysis. It is simply a fact that only the data for 2008 is missing. The onus is on PP to produce either the data or the reason why the data has not yet been produced.

You explicitly said the data was withheld (not missing), and theorized that it was to hide a spike in abortions. You follow it up by explaining that the missing data in the chart must be missing because PP didn't provide it.

Found the report: https://www.yumpu.com/en/document/view/14436304/ppfa-annual-...

Supports up to 25% YOY change from 2007. Can you find this report on PP's site? Or the breakdown of the absolute values?

And yet the report from Politifact in QZ that explicitly shows "year over year change" shows no such thing:

- 5% in 2007

- 9% in 2009

- -1% in 2010

- 2% in 2011

- -2% in 2012

- 0% in 2013

Please read the thread. You missed the 2008 report (as did politifact). I linked to the 2008 report which does not show absolute service number breakdowns; only rounded (or, perhaps, truncated [in which case the YOY change could be significantly higher than 25%]) percentages from a total.

Why would PP deviate from their standard reporting practices?

I did read the thread. You quoted a report that showed '25% year over year'. With the numbers of this report, there would have to be a 100% increase in 2008 (that reverted back to "normal" levels in 2009) to average out to that level.

Even rounding numbers would not account for the discrepancy between "2% more this year than last" and "25% more this year than last".

Its not percent of a percent. It's percent of an absolute change.

1st derivative vs 2nd derivative.

Edit: hmm, maybe I misunderstood your point. I don't think the politifact or qz YOY data from 2009 is based on 2008 (but rather, 2007 [technically not YOY then]). How would they have calculated the YOY without the absolute number for 2008?

Why is there an onus on them to do anything? They don't owe you any figures on how many legal medical procedures they do with money that didn't come from you.

Because they have the data. No one else does.

Again, how does that oblige them to release to you, or anyone, such data?

The only obligation would be to have a complete organizational record. As it stands, only 2008 has an incomplete record. Why do you think that would be the case?

Reply on news.ycombinator.com