Like many Americans during a presidential election year, I am naturally interested in poll results. While I do not like to discuss my political views in public (and please do not interpret any of my statements in this article as such), I do like to discuss data visualizations. Especially bad ones.

像总统选举年期间的许多美国人一样，我自然对民意调查结果感兴趣。 虽然我不想公开讨论我的政治观点(并且请不要这样解释我在本文中的任何陈述)，但我还是想讨论数据可视化。 特别是不好的。

For example, this visualization from Ipsos, showing the results of an ABC News / Ipsos poll published on July 31. Pop quiz: what is wrong with it?

例如，来自益普索(Ipsos)的此可视化效果，显示了7月31日发布的ABC新闻/益普索(Ipsos)民意调查的结果。

1.环图 (1. Ring chart)

For those of you who have not encountered it before, the pie chart variant shown above is known as a ring chart or a sunburst chart. There has a been a general trend away from using pie charts because studies have shown that people are not nearly as good at comparing the size of angles as they are at comparing length. In other words, using a pie chart at all can make it too easy for the reader to misinterpret the data. In this case, a simple bar graph would have been intuitive and sufficient.

对于以前从未遇到过的人来说，上面显示的饼图变体被称为环形图或朝阳图。 不再使用饼图是一个大趋势，因为研究表明，人们在比较角度大小方面不如在比较长度方面好。 换句话说，完全使用饼图会使读者很容易误解数据。 在这种情况下，简单的条形图将是直观且足够的。

Furthermore, this is a completely improper use of a ring chart. Ring charts are meant to display hierarchical data, which does not — for example, Democrats are not a subset of Republicans. A standard ring chart looks like the one below (courtesy of FusionCharts.com). Notice how the inner ring shows the broadest categories of continents, the middle ring divides those into regions and the outer ring further divides those regions into individual countries.

此外，这完全是对环图的不当使用。 环形图旨在显示分层数据，而不显示分层数据，例如，民主党不是共和党的子集。 一个标准的环形图表如下所示(由FusionCharts.com提供 )。 请注意，内圈如何显示最广泛的大陆类别，中圈将其划分为多个区域，外圈进一步将这些区域划分为各个国家。

Had the authors of the Ipsos chart wanted to use a ring chart correctly, they could have shown an inner ring representing Americans who approve versus Americans who disapprove, and an outer ring breaking those groups down further into Democrats and Republicans. Or, the inner ring could represent Democrats versus Republicans, and the outer ring could break those groups down into Americans who approve versus Americans who disapprove.

如果Ipsos图表的作者希望正确使用圆环图，他们本可以显示一个内圈，代表赞成的美国人与不赞成的美国人，以及一个将这些群体进一步细分为民主党和共和党的外圈。 或者，内圈可以代表民主党与共和党的对抗，而外圈可以将这些群体分为赞成与反对的美国人。

Had they done this, it would have revealed something hidden by the current chart: the missing subset.

如果执行此操作，则可能会显示当前图表隐藏的内容：缺少的子集。

2.缺少子集 (2. Missing subset)

The colored sections of the Ipsos ring chart show the “percentage of Americans who approve of the way Donald Trump is handling the response to protests happening across the country.” It appears that we are meant to assume that the large swaths of glaring white space represent those who disapprove, and it turns out that assumption would be more or less correct. A look at the survey results shows that less than 0.5 percent of respondents skipped the question. In general, however, including the results for “skipped” or “no opinion” is a good practice in order to limit the reader’s assumptions. Suppose 8 percent of Democrats approved of the President’s actions, 8 percent disapproved and 84 percent just didn’t care — in this election cycle, that level of apathy would be a newsworthy story, but we’d have no way to know from this chart.

益普索圆环图的彩色部分显示了“ 赞成唐纳德·特朗普处理全国各地抗议活动的方式的美国人比例。” 看来，我们本来是要假设一大堆耀眼的空白代表那些不同意的人，事实证明，这种假设或多或少是正确的。 查看调查结果表明，不到0.5％的受访者跳过了该问题。 但是，总的来说，包括“跳过”或“没有意见”的结果是一种很好的做法，以限制读者的假设。 假设有8％的民主党人赞成总统的行动，有8％的人反对，而84％的人不在乎-在这个选举周期中，冷漠的程度是一个值得新闻报道的故事，但我们无法从这张图表中得知。

However, “no opinion” is not the missing subset I’m talking about. 8 percent of Democrats approve of the President’s actions. 78 percent of Republicans approve. What about the independents?

但是，“无意见”并不是我正在谈论的缺失子集。 8％的民主党人赞成总统的行动。 78％的共和党人赞成。 那独立人士呢？

According to the most recent Gallup poll, taken between July 1 and July 23, a plurality of Americans — 38 percent! — do not affiliate with either major political party. The survey methodology for the Ipsos poll indicates that they considered party identification (“Democrat, Republican, Independent, Other/None”) in their weighting scheme. So why are only Democrats and Republicans shown on the chart? Isn’t a party breakdown meaningless if you exclude the largest group?

根据盖洛普(Gallup) 7月1日至7月23日进行的最新民意测验 ，有38％的美国人是美国人！ -不隶属任何主要政党。 益普索民意调查的调查方法表明，他们在权重方案中考虑了政党身份(“民主党，共和党，独立，其他/无”)。 那么，为什么图表上只显示了民主党人和共和党人呢？ 如果排除最多的团体，聚会破裂不是没有意义吗？

Side note: the methodology section of the Ipsos poll indicates that the results were weighted to adjust for various demographics such as gender, age and ethnicity, as determined by a recent U.S. Census Bureau survey, the results of which are publicly available. They were also weighted to adjust for party affiliation proportions as determined by recent ABC News / Washington Post telephone polls — but nowhere does the methodology section state what those proportions were determined to be. So if, for example, Republicans refuse to answer phone calls from ABC News or the Washington Post in the first place, they could infer that the Republican proportion of the population is, say, only 1 percent. High Republican approval of the President’s actions would then have a negligible impact on the overall approval rating. This would be an important thing for the reader to know when assessing the reliability of the analysis.

旁注：益普索(Ipsos)民意调查的方法部分指出，加权结果是根据美国人口普查局最近的一项调查确定的，以适应诸如性别，年龄和种族的各种人口统计数据，其结果可公开获得。 根据最近的ABC新闻/华盛顿邮报电话民意测验，他们也被加权以调整党派的比例-但是方法论部分没有任何地方说明这些比例被确定为什么。 因此，例如，如果共和党人首先拒绝接听ABC新闻或《华盛顿邮报》的电话，他们就可以推断出共和党人的比例仅为1％。 这样，共和党对总统行动的高度支持将对总体支持评级产生微不足道的影响。 这对于读者在评估分析的可靠性时要知道的一件重要事情。

3.配色方案 (3. Color scheme)

Since the 1980s, Democrats have consistently been associated with the color blue, and Republicans with the color red, so it makes sense to use those colors to represent those groups on this chart. There is no valid reason to represent “All Americans” with a similar shade of dark blue as Democrats. For one thing, it’s visually confusing. The first group shown in the color legend is “All Americans,” and right next to it, the outermost ring of the chart shows “8%.”

自1980年代以来，民主党人一直将蓝色与红色联系起来，共和党人将红色与红色联系在一起，因此在图表上使用这些颜色来代表这些群体是有意义的。 没有充分的理由用与民主党人相似的深蓝色来代表“所有美国人”。 一方面，它在视觉上令人困惑。 颜色图例中显示的第一组是“所有美国人”，在其旁边，图表的最外面的环显示为“ 8％”。

A reader could easily jump to the wrong conclusion that only 8 percent of all Americans approve of the President’s actions, not 8 percent of Democrats.

读者很容易得出一个错误的结论，即只有8％的美国人赞成总统的行动，而不是8％的民主党人。

4.斜体 (4. Italics)

This is a minor thing, but it bugs me. The title of the visualization is “Percentage of Americans who approve of the way Donald Trump is handling the response to protests happening across the country.” Why is “approve” in italics? Given the pattern of the other issues I’ve noted with this visualization, the only reason I can think of is editorial bias. It reads like they assume the default state is disapproval, and the fact that anyone at all approves of the President’s actions is surprising. There’s no need for the authors to inject personal opinion here; a good visualization lets the numbers tell the story.

这是小事，但令我感到烦恼。 可视化的标题是“ 赞成唐纳德·特朗普处理全国各地抗议活动的方式的美国人所占的百分比。” 为什么以“斜体”表示“赞成”？ 考虑到我在此可视化中注意到的其他问题的模式，我能想到的唯一原因是编辑偏见。 这听起来像是他们假设默认状态不被接受，任何人都认可总统的行为的事实令人惊讶。 作者无需在此处插入个人意见。 良好的可视化效果可以使数字说明问题。

Do you see any other potential problems with this visualization? Am I being overly critical? Let me know your thoughts!

您还发现此可视化还有其他潜在问题吗？ 我是否过于挑剔？ 让我知道你的想法！

Images in this article may include copyrighted material, the use of which has not been specifically authorized by the copyright owner. Under Title 17 U.S.C. Section 107, allowance is made for “fair use” for purposes such as criticism, comment, news reporting, teaching, scholarship, education and research.

本文中的图像可能包含受版权保护的材料，未经版权所有者明确授权而使用该材料。 根据《美国法典》第17条第107款，为“合理使用”提供了津贴，用于批评，评论，新闻报道，教学，奖学金，教育和研究等目的。