follow CCP

Recent blog entries
popular papers

Science Curiosity and Political Information Processing

What Is the "Science of Science Communication"?

Climate-Science Communication and the Measurement Problem

Ideology, Motivated Cognition, and Cognitive Reflection: An Experimental Study

'Ideology' or 'Situation Sense'? An Experimental Investigation of Motivated Reasoning and Professional Judgment

A Risky Science Communication Environment for Vaccines

Motivated Numeracy and Enlightened Self-Government

Making Climate Science Communication Evidence-based—All the Way Down 

Neutral Principles, Motivated Cognition, and Some Problems for Constitutional Law 

Cultural Cognition of Scientific Consensus

The Tragedy of the Risk-Perception Commons: Science Literacy and Climate Change

"They Saw a Protest": Cognitive Illiberalism and the Speech-Conduct Distinction 

Geoengineering and the Science Communication Environment: a Cross-Cultural Experiment

Fixing the Communications Failure

Why We Are Poles Apart on Climate Change

The Cognitively Illiberal State 

Who Fears the HPV Vaccine, Who Doesn't, and Why? An Experimental Study

Cultural Cognition of the Risks and Benefits of Nanotechnology

Whose Eyes Are You Going to Believe? An Empirical Examination of Scott v. Harris

Cultural Cognition and Public Policy

Culture, Cognition, and Consent: Who Perceives What, and Why, in "Acquaintance Rape" Cases

Culture and Identity-Protective Cognition: Explaining the White Male Effect

Fear of Democracy: A Cultural Evaluation of Sunstein on Risk

Cultural Cognition as a Conception of the Cultural Theory of Risk

« Midweek update: teaching criminal law--voluntary manslaughter | Main | Next stop (not counting weekly trips to Cambridge, MA) »

Weekend update: does transparency help with this overplotting problem?

Another example of how to use transparency functionality of Stata 15.

Compare this ...

 ... with this:

Which one is better? Why? Other ideas?


PrintView Printer Friendly Version

EmailEmail Article to Friend

Reader Comments (37)

Given the number of data points you have, I think both look pretty good. Though, in my opinion the one without transparency actually allows you to get a better handle on the number of points in an area.

That said, I really prefer a 2d histogram (with color being the height) for these types of plots. Here is a Python example. It helps show exactly how many points are in a given area.

November 5, 2017 | Unregistered CommenterZach Hafen

I agree with Zack Hafen that there are graphical methods that better describe the data.

But I also think that it is high time that some other scale than the x axis shown in all of these articles needs to be changed into something with more dimensionality.

Nearly 1/3 of the US registered electorate identify as Independent, even though, for the purposes of voting, they may "lean" one way or the other given the unofficial, but still locked in constraints of our 2 party system. It is not necessarily where they really are, ideologically, but it is where the available candidates are. Thus it does not mean that a D or R choice accurately reflects their value system. Nor does it clearly define those who do register with party identification. Trump is not Romney or Jerry Falwell, Clinton is not Sanders or Depok Chopra. Both parties are underlain by funders seeped in Wall Street. More dimensionality, including nodes of commonality are needed. Additionally, large numbers of citizens are not registered to vote at all.

Additionally some of the things that one might assume one would want to have controlled by the nation state, of which climate change and air pollution would be prime examples, are in fact global issues with only global solutions. We have no authoritative structures for dealing with that.

Additionally, our information systems have globalized, such that propaganda and culturally cognitive based manipulation now are coming from abroad into local matters to a much greater extent.

Corporations have long been global in reach: look at the role of the East India Company and Hudson's Bay, in British colonialism. But many large companies now are transnational, operating without regard for a "home: nation. If currency alternatives like bitcoin take place, financial transactions and thus commerce can increasingly take place outside of the structures of nations and central bankers.

Information now gives us new global actors, the net-states: This ought to be a focus of cultural cognitive research, as it is central to our future:

"But deploying traditional military tactics in battles of belief are the equivalent of setting bear traps for ghosts: They’re not going to work. They’re not relying on the wrong weapons; they’re relying on the wrong worldview."

The first baby step forward, IMHO, would be in coming up with depictions of the US public world view that are not locked in to a linear R to D scale.

November 5, 2017 | Unregistered CommenterGaythia Weis

@Zach, thanks. We've tried various 3d renderings of this iconic graphic w/o being satisfied. But this might work better --thx

November 5, 2017 | Registered CommenterDan Kahan

When it comes to climate change: some dimension that includes falling to the far right on the above scale that does not explicitly involve identifying as a strong Republican seems to be in operation. From a study by some other Yalies:

"Earlier this year, several prominent Republicans, including James Baker III (former Secretary of State, Treasury, and two-time White House Chief of Staff under Presidents Reagan and George H.W. Bush), George Schultz (former Secretary of State and Treasury under President Reagan), and Henry Paulson Jr. (former Secretary of Treasury under President George W. Bush), called for a national carbon tax with all revenues returned to taxpayers as dividends."

Politically, anti-climate change advocacy could have a lot to do with acquiring the necessary financial backing to currently become an elected politician. I'd put my (Colorado) US Senator Cory Gardner and US Representative Ken Buck in that category. Along with that would be the wherewithal to utilize the sorts of culturally cognitive bases strategies of data mining and "strategic communication" to "change audience behavior" as employed by corporations such as Cambridge Analytics:

Out of the confined box of "lesser of two evils" voting strategies, provided by our current embedded two party structure, what would people do? From the point of view of policy change likely to be influenced by the next election, this may not be a consequential question. But that should not be confused with a fundamental attribute of peoples actual attitudes.

Which does offer the hope that with the right information access and positioning, perspectives and priorities could be changed, and alternative policies implemented.

But a billion dollar or at least multi million dollar budget would be helpful.

November 5, 2017 | Unregistered CommenterGaythia Weis

A few thoughts...

You possibly want the boundaries of your semi-transparent circles to be semi-transparent too.

Have you considered doing a density plot of the 'kernel density estimate'? (see the last plot here: )

The 'jitter' plots where discrete data points have random noise added so you can distinguish them work so long as the density is not too high, but you lose precision with overplotting. Perhaps a better approach is to draw 8 very narrow bar plots (or violin plots) so that the thickness is exactly proportional to the change in density. Something like this

November 5, 2017 | Unregistered CommenterNiV

@Gaythia-- you know that political outlooks are being used only asn indicator of some disposition we can't observe directly. It's value lies not in sociological realism of any sort but only its relative ease of use and its satisfactory power to explain & predict. If you prefer another proxy, fine, but spell out what sorts of observations it comprises and why they are better than left-right. I could give such an account for using cultural worldviews instead of left-right. But left-right explains the variance well enough, is more readily understood b/c more familiar to most people, and is easier to work with b/c it is 1 dimensional rather than 2. So you would need to tell me what indicators to use for your preferred alteratnive, how to measure thoes indicators, and why the would enhance understanding to a greater degree.

I don't doubt you could, but I don't really feel you have yet.

November 5, 2017 | Registered CommenterDan Kahan

@NiV-- thanks. Maybe I should post the data & ask for people to do what they are recommending and supplying the code for how they did it

November 5, 2017 | Registered CommenterDan Kahan

link drop - the power of optimistic anecdotes in science communication:

November 6, 2017 | Unregistered CommenterJonathan

@jJonathan-- good. I'm glad the controversy will soon be over

November 6, 2017 | Registered CommenterDan Kahan

Dan -

I thought that making smart people more curious about science will end the controversy.

November 6, 2017 | Unregistered CommenterJoshua


Not one of your better efforts. However, seriously, here's an alternative to preaching positive anecdotes:

Focus on the lay public’s trust of the scientific community (in cases where there is a strong consensus), rather than on individual scientists. It is one thing to be able to recognize cases of scientific consensus. It is quite another to recognize the epistemic significance of such consensus. Why is it that this consensus should interest us? What kind of consensus is important? What is it about the scientific community makes this so? We submit that these are matters for which that public’s understanding of science could be improved. The suggestion is that improving them may result in greater willingness to seek and defer to scientific consensus where it exists.

November 6, 2017 | Unregistered CommenterJonathan

"@NiV-- thanks. Maybe I should post the data & ask for people to do what they are recommending and supplying the code for how they did it"

Well, I don't have Stata, so I can't help you there!

But here's how to do it in R.

# --------------------------------------------------------------------------------
# Make up some data
y = floor(runif(2000,0,8)) # Uniform Random in interval 0 to 7
x = rbinom(2000,5,1-(y+1)/9) + rnorm(2000)/4 - 2.5 # Jittered binomial to give smooth skewed distributions
d = data.frame(Risk=factor(y),LeftRight=x) # Stick them together in a data frame for ggplot2
title = "\"How much risk do you believe global warming\nposes to human health,safety, or prosperity?\""

# Scatter plot with circles
plot(x,y+rnorm(2000)/10,col=rainbow(8)[y+1],xlab="Left-Right",ylab="Risk Assessment",main=title)

# Scatter plot with semi-tranparent discs
plot(x,y+rnorm(2000)/10,cex=2,pch=16,col=rainbow(8,alpha=0.1)[y+1],xlab="Left-Right",ylab="Risk Assessment",main=title)

# Kernel density estimate
dens <- kde2d(x,y,n=100)
image(dens,xlab="Left-Right",ylab="Risk Assessment",main=title)
contour(dens, add=T)

# Violin plot
ggplot(d, aes(x=Risk, y=LeftRight,fill=Risk)) + coord_flip() + geom_violin() +
ggtitle(title) + theme(plot.title = element_text(size=14, face="bold",hjust=0.5))
# --------------------------------------------------------------------------------

I've not spent a lot of time adding polish, proper axis labeling and so on, so you could get something a lot better with more effort. But they give the rough idea, and a starting point for experimentation.

This one's the violin plot, for example.

November 6, 2017 | Unregistered CommenterNiV


The basic problem with all these forms of advertising/brainwashing/Arguments-From-Authority is that *both* sides of any controversy can use them with equal facility. There's nothing in the technique for distinguishing truth from lie.

If you lie to people and tell them that using mobile phones causes brain cancer, because 97% of scientists say so, then show them some viscerally emotional imagery to emphasize the message, people will use mobile phones less. That's not a good thing.

November 6, 2017 | Unregistered CommenterNiV


"The basic problem with all these forms of advertising/brainwashing/Arguments-From-Authority is that *both* sides of any controversy can use them with equal facility."

I'm well aware of that. The problem is: what manner of science communication both addresses how to distinguish truth from nontruth *AND* has a positive impact on people's beliefs and choices. Anecdotes are well known to impact people's beliefs and choices, but don't carry much scientific water (and so can be weaponized easily). Actually - I'd add a 3rd branch to the above conjunction: *AND* respects cognitive dignity - hence not usage of biases. It's one big tough nut to crack, and might simply not be crackable. What do to then?

November 6, 2017 | Unregistered CommenterJonathan

"What do to then?"

Teach people that in science "I don't know" is always an acceptable answer.

From a scientific point of view, the alternatives are to teach people to use the scientific method, or to teach them to say "I don't know". And to teach them that the people selling them any other message - especially Arguments from Authority - are not good scientists.

It's all about the uncertainty.

November 6, 2017 | Unregistered CommenterNiV

"Teach people that in science "I don't know" is always an acceptable answer."

This has the unfortunate side-effect of contributing to their status-quo bias. Hence it is also necessary to teach them when and how "we know enough" to justify disrupting the status-quo.

November 6, 2017 | Unregistered CommenterJonathan

Dan - is there any way to extend the y-axis into the negative by adding the option "global warming has a positive effect"?

You would get few respondents picking that option, so quantifying it is probably superfluous. But that you would get at least some is certain, because the "risk of global warming" is overwhelmingly rejected by the alt-right. It would be similar to the Gamergate backlash against a developer named Zoe Quinn, who tried to promote her game on Depression and got mercilessly hounded online. This is a good summary of the event, for those unfamiliar:

"......“Gamergate”, a misogynistic campaign in which mostly white men keen to bolster each other’s egos let rip against feminists and all the other “social justice warriors” they despised in the world of gaming and beyond. According to some estimates, more than 2m messages with the hashtag #gamergate were sent in September and October 2014. [....} ........ the mainstream media, when they noticed it, misinterpreted it. They took Gamergate to be a serious debate, in which both sides deserved to be heard, rather than a right-wing bullying campaign......"

The debate was, in fact, viewed as a serious one by the alt-right - it is the article's author who decided to label it "bullying". And this is analogous, and possibly completely symmetrical, with the alt-right's view of the "97% consensus" being a SJW bullying campaign. The axes as they presently stand ignore that portion of the N=2,000.

November 7, 2017 | Unregistered CommenterEcoute Sauvage

PS I like the first graph better - it's less fuzzy visually. And this is more on Gamergate - and how everything politically correct is viewed by the alt-right as against "free speech".

"........... The argument shifted away from video games to how “social-justice warriors” like Quinn and her defenders (in other words, anyone who suggested the harassment was caused by sexism and prejudice) were intent on bending the culture to their politically correct will. Free speech — even masculinity in general — was under attack......"

Masculinity did not play much of a role, as I recall, (many women, myself included, thought Quinn is a kook, not to mention that Gamergate was originally publicized by Milo Yannopoulos) so this is just the article's author trying to introduce sexism into the equation.

November 7, 2017 | Unregistered CommenterEcoute Sauvage

"This has the unfortunate side-effect of contributing to their status-quo bias."

Much to the disappointment, I'm sure, of all those people who want to disrupt the status quo without the decision makers having any solid knowledge of whether their proposals are a good idea or not.

"Hence it is also necessary to teach them when and how "we know enough" to justify disrupting the status-quo."

It's the "how" that's at issue here. If your alternatives are the scientific method or not knowing, then obviously for people to know when they know enough, they have to be taught to apply the scientific method.

Again, the problem is that the alternative approach can be used to justify *any* policy, including the exact opposite of the one being proposed. You either use a scientific approach or you don't.

November 7, 2017 | Unregistered CommenterNiV

Jonathan -

You might like this one:

November 7, 2017 | Unregistered CommenterJoshua

Oh - sad - I was rooting for the ego depletion result. Were I Dan, I'd lose self-control and kick my (his) cat.

Anyway, at least the cream reduces the skin rash:

non-paywall version of paper:

November 7, 2017 | Unregistered CommenterJonathan

Re polarization: factors that have nothing to do with scientific curiosity, cognitive reasoning, or even political ideology:

November 8, 2017 | Unregistered CommenterJoshua

And an interesting twist w/r/t political ideology and climate change:

Trump won 27 of the 32 states, almost entirely across the nation’s interior, with the highest per capita levels of the carbon dioxide emissions linked to global climate change. Clinton won 15 of the 18 states with the lowest per capita carbon emissions –

November 8, 2017 | Unregistered CommenterJoshua

From that article:

, Trump has lost support in office from both sides of his coalition: His approval rating today in surveys is consistently lower than his share of the vote last November among both blue-collar and older whites (especially women) and white-collar whites.

Although national polling in the 2016 election - regarding voting intentions - was actually pretty accurate (despite claims otherwise by political hacks) it might be a mistake to compare polling support for job performance or favorability ratings, to votes.

That said, I do believe that he has lost polling support for job performance and favorability among all those groups as well, although to varying degrees.

November 8, 2017 | Unregistered CommenterJoshua


Lots of potential good (and free!) reads here, as the British Journal of Sociology celebrates the first anniversary of Trump (and Brexit):

November 8, 2017 | Unregistered CommenterJonathan

Thanks Jonathan.

November 8, 2017 | Unregistered CommenterJoshua

An article that provides (anecdotal) evidence that the relationship between ideological identification and issues and "values" may be quite complicated.

So the issues that many said motivated their votes for Trump matter less now that he's in office?


November 9, 2017 | Unregistered CommenterJoshua

It’s not that the people who made Trump president have generously moved the goalposts for him. It’s that they have eliminated the goalposts altogether.

November 9, 2017 | Unregistered CommenterJoshua

Ok. Thus is perfect:

“Everybody I talk to,” he said, “realizes it’s not Trump who’s dragging his feet. Trump’s probably the most diligent, hardest-working president we’ve ever had in our lifetimes. It’s not like he sleeps in till noon and goes golfing every weekend, like the last president did.”

I stopped him, informing him that, yes, Barack Obama liked to golf, but Trump in fact does golf a lot, too—more, in fact.

Del Signore was surprised to hear this.

“Does he?” he said.

“Yes,” I said.

He did not linger on this topic, smiling and changing the subject with a quip. “If I was married to his wife,” Del Signore said, “I don’t think I’d go anywhere.”

Know that there's some kickass knowing disbelief.

November 9, 2017 | Unregistered CommenterJoshua

I think that the Ronald Brownstein analysis given by Joshua above at this link: ttp://, actually has quite a bit to do with elucidating the identity and cultural cognitive drivers that determine how people assimilate information related to political policy. And how they vote, and change their votes.

All of which is highly relevant to our current election season.

To being to answer Dan, I think that for starters we have to recognize that the Cultural Cognition project, and it's efforts to inform science communication work in general does have ulterior motives, at least some of which his website has given as follows:

“This project seeks to use science's own signature methods of disciplined observation and inference to identify and solve the "science communication problem" distinctive of contemporary pluralistic democracies.”

“The Project also has an explicit normative objective: to identify processes of democratic decision-making by which society can resolve culturally grounded differences in belief in a manner that is both congenial to persons of diverse cultural outlooks and consistent with sound public policymaking.”

In the post above, Dan notes:
“political outlooks are being used only as an indicator of some disposition we can't observe directly”

I think that it is key to note that the problem with this approach is that “political outlooks” as arranged above on a linear basis isn't a measure of “some disposition” at all, but rather an amalgam of dispositions.

Who ends up voting for a Democrat or a Republican is determined by how individuals see, and are led to see, their own, frequently multiple, identities best expressed. This often turns out to be voting for what are quite often regarded as a “lesser of two evils” candidate. I would postulate that identification as a Democrat or Republican is not overriding for most Americans. Certainly not for the “swing” and “reluctant” voters who determine the fate of election outcomes.

In my opinion, it is the narrowing to a dichotomy, as force by our two party political process, that limits constructive, consensus building debate. This is not intrinsic to human civilization nor to the democratic process. It probably did flow naturally out of situations in which a single issue seemed to be the deciding matter in a given election. But then the parties got bureaucratized and “stuck' with new issues that came along not forming new affiliations among like minded voters, but rather being presented by the parties as part of a package designed to fit with preconceived identities.

As a different and contrasting example, in the local election that took place yesterday, many of the Colorado City Council elections conducted elections in which voters cast votes for multiple candidates. Candidates were to be “nonpartisan” and political party identification was not part of the campaign process. In one nearby community, voters could cast their ballots for up to 5 candidates. This leads to a largely different governing body, one which is much less likely to be winner take all or sharply divided on opposite sides of one line.

In the graph in Dan's post above, the x axis in the graph doesn't offer much information as to how cultural identities are used by individuals, and exploited by political consultants, to end up with the winning amalgam that can win elections or create support for new policy positions. IMHO, it obscures it.

The culturally cognitively meaningful data is buried. Since the political parties have platforms and agendas upon which candidates who wish to obtain funding need to make “litmus test” positions, the information requested on the y axis is actually also a confounding variable of the x axis. As noted in my comment above, for any given point on the x axis, there are a number of very different sorts of people that might be placed there. It doesn't seem to me to tell us much about how to communicate effectively with them. Nor does there seem to me to be much statistical significance or graphical representational significance in the details of how this x-y plot is expressed.

Given the difficulties of coming anywhere close to the data mining available to a Facebook or a Cambridge Analytics, I think that the way forward for academia would be in carefully constructed analysis of clearly defined segments of the total population.

As a study of effective exploitation of identity and emotion to drive responses, I think that it is fascinating to observe how kneeling, an action formerly seen as one of respect or even submission, has effectively been turned into one that some people, some of whom are perhaps more likely to kneel in a religious fashion than generally, are now seeing as intensely disrespectful.

I've also read Joshua's latest link above, I think we should go back and re-study history with a take giving more empathy to the Luddites. We haven't given many of the remaining citizens of places like Johnstown Ohio a clear path forward. They are strongly on the losing end of the disappearance of the middle class. The ancestors of Luddites may have been poor, but until mechanization displaced them on the farm, they were arguably living in better surroundings in their poverty than the first workers in a polluted mill town. It took unionization to wrest benefits of industrialization back from the robber baron overlords to the workers. Something similar is needed now.

November 9, 2017 | Unregistered CommenterGaythia Weis

Culturally cognitive informed mind hijacking:

November 9, 2017 | Unregistered CommenterGaythia Weis

Gaythia -

Do you have more info on that Colorado City Council elections? Looks interesting. I could Google, but you might offer a link from an informed perspective.

November 9, 2017 | Unregistered CommenterJoshua

My city, Longmont, has a mixed system. I voted for one of two candidates for my "Ward", a regional designation. Longmont has 3 Wards. I also voted for two "at large" candidates. These positions were won by the top two vote getters out of a list of 5 candidates. city wide.

I thought it was 5, but it appears that the nearby City of Lafayette voted for 4 candidates out of a list of 14. The results for that are described here:

For voting, Boulder county uses a mail in paper ballot that is tracked and scanned. This paper trail means that the results can be later recounted and verified:

I find voting methods to be an interesting topic for potential electoral reform. There are methods which give a fuller representation of overall voter positions, but all have some pitfalls.

I've been somewhat involved with a local League of Woman Voters committee that has been looking at alternative methods including "instant runoff" or "transferable vote". All of this is more readily possible with automated methods of vote tallying.



November 9, 2017 | Unregistered CommenterGaythia Weis

Thanks Gaythia -

A follow on to that other article:

November 10, 2017 | Unregistered CommenterJoshua

Gaythia - if you are looking for a cultural v. left-right scale, this is an international one:

As to your problems with the x-y axes above, my main problem with them is that they really are a sequence of x-y-z plots (intensity of color denoting the z-scale) but 1. give no indication of portion of the N=2,000 in each mini-distribution, and so therefore no conclusion on the cumulative x-y-z distribution can be obtained from the partial ones shown, and, 2., my own group (that more CO2 would be a welcome addition to the atmosphere, and that measures to limit CO2 emissions create more pollution) isn't even represented.

To see some of the more spectacular disasters of the limit-CO2-emissions crowd, consider the folly of substituting diesel-powered cars for gasoline-burning ones (a main point of the Kyoto protocol) and, more recently, the electric/hybrid car craze, substituting lithium-cobalt (powered by coal power generation) for the old car models.
Shutting down nuclear power generation in favor of burning woodchips, building windmills, installing solar panels is too absurd to even comment on.

Finally, the latest, and funniest, geoengineering CO2-decreasing schemes, stand in a class of their own. This from a professor Keith of Harvard:

"...........the House Science Committee, whose chair recently called concern over climate change “hysteria,” will hold a geoengineering-focused hearing. Keith frets that such hearings will backfire.

“There are some real potential downsides to rapid promotion of these technologies by the administration we have,” he said. Those include disrupting the fragile political coalitions backing policies for emissions reduction. “In some ways the thing we fear the most is a tweet from Trump saying, Solar geoengineering solves everything! It’s great! We don’t need to bother to cut emissions.”............."

Actually, prof. Keith is right, because that is the exact move now planned by the administration :)

November 11, 2017 | Unregistered CommenterEcoute Sauvage

PS link to prof. David Keith speech

November 11, 2017 | Unregistered CommenterEcoute Sauvage

Like the transparency; makes it easier to follow the density in the distribution as it changes from top to bottom. Don't know if others are seeing the same effect, or if the effect itself is misleading in some strange way...

Transparency makes me like the line much less.

November 30, 2017 | Unregistered Commenterdypoon

PostPost a New Comment

Enter your information below to add a new comment.

My response is on my own website »
Author Email (optional):
Author URL (optional):
Some HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>