Pollster.com

Articles and Analysis

 

New Charts for Polls, Polls, Polls

Topics: Charts , Pollster.com , Trend lines

A little over two years ago, we launched Pollster.com with a mission of providing a complete compilation of poll results, expert analysis and graphical tools to help readers make sense of polling data. Today, after two long years of development, our commitment to interactive graphical tools takes a quantum leap.


At a moment when the political world is swimming in a flood of polling data, we are pleased to announce a new, fully interactive Flash chart application that will plot all of the poll charts here on Pollster.com. The new charts allow you to:

  • Select or limit the polls used to draw trend lines and calculate polling estimates with the "filter" tool. If you don't like a particular pollster, just un-click and take them out (yes...really).
  • Toggle between the display of the default trend line and alternatives that are more or less sensitive using the "smoothing" tool -- these are essentially the same as the "steady blue" and "ready red" trend lines often used by Charles Franklin.
  • Hold your mouse over any data point to display details about each the poll.
  • Click the mouse on any data point to "connect the dots" between all polls fielded by that pollster.
  • Modify the date range (x-axis) and percentage range (y-axis) by clicking on either axis directly or with forms found on the "tools" menu.
  • Select the candidates you want to see displayed on the chart with the "choices" tool .
  • Toggle the display of data points, trend lines and grid lines on or off with the "plot" tools.
  • Copy the code necessary to bookmark your customized chart or share it via email with the "URL" tool.
  • Get the code necessary to place a small version of the customized chart on your own blog or web site with the "Embed" tool.  [Update: We believe we've squashed the embed bug.  If you experience problems with the embed tool or anything else, please email us with details at questions at pollster dot com.  We missed a bugin the embed function that prevents the embedded chart from displaying customized filters.  Apologies -- we should have this cleaned up soon]

As of this posting, we have converted our charts for all presidential charts, including Ohio, Colorado, Virginia , Florida, Michigan, Pennsylvania, New Hampshire, Nevada, Minnesota and the National Trend to the new chart, and we should have most of the presidential charts converted over the course of the day. Although we hope you will dive right in and try these new features, we have also created a quick video guided tour of some of the most important features.




We are really excited about the possibilities these new tools create for poll junkies to explore and discover the wealth of data now available and how you are too. So if you like If you like what you see, we hope that you share this news with your friends. If you have a blog or diary, we would very much appreciate a link to this entry and please try out the embedding feature to see how it works. I'll be back later with some extra tips on how to use the charts. Meanwhile, please don't hesitate to email us with your questions or reactions.

Where credit is due: A lot of very talented people worked exceptionally hard to make these new charts possible, but most deserving of thanks are Quentin Fountain and Technorganix (for Flash design and development), Jeff Lewis and Seth Hill (for development of the underlying database and statistical architecture), Charles Franklin (for the design of the original charts and the regression trend lines and smoothing routines) and, last but not least, Eric Dienstfrey, who has entered virtually every piece of the data now displayed on Pollster.com and is doing the nitty gritty work of implementing these charts and keeping Pollster operating every day.  Update: I nearly forgot to thank those of you -- and you know who you are -- who helped beta test these charts over the past week.

 

Comments
Allen:

To follow up on the recent discussion of "coloring" states, I did a comparison between the results at Pollster.com and election-projection.net. While the coloring is admittedly arbitrary, in all cases where it could be questioned, I think election-projection has it right, and Pollster has it wrong.

To summarize the methodology differences: Pollster colors based on the confidence of the margin, where the two candidates scores come from the LOESS projections, and the margin of error comes from the average poll sample size.

election-projection.net does not color but instead reports probabilities from 0 to 100%. It computes a weighted average of recent polls, with the weights based on sample size and poll age. The margin of error is the larger of the sampling error or the actual poll variance, plus a constant of 1.75%. See href="http://election-projection.net/mathematics.html">election-projection.net methodology.

election-projection.net has ME at 91% Obama based on the 4 most recent polls that give Obama an average margin of 7.8%. Pollster has ME leaning Obama. I think it should be solid Obama.

election-projection.net has WA at 91% Obama based on the 6 most recent polls that give Obama an average margin of 6.2%. Pollster has WA leaning Obama. I think it should be solid Obama.

election-projection.net has CO at 90% Obama based on the 6 most recent polls that give Obama an average margin of 5.6%. Pollster has CO a loss up. IMO based on recent polling alone, CO should be at minimum lean Obama, if not solid Obama.

election-projection.net has PA at 83% Obama based on the 4 most recent polls that give Obama an average margin of 4.0%. Pollster has PA a toss up. I think it should be lean Obama.

election-projection.net has LA at 92% McCain based on the 3 most recent polls that give MCain an average margin of 9.9%. Pollster has LA lean McCain. I think it should be solid McCain.

election-projection.net has MT at 93% McCain based on the 3 most recent polls that give MCain an average margin of 8.6%. Pollster has MT lean McCain. IMO based on recent polling alone, MT should be colored solid McCain.

Those are the major discrepancies. It would appear that Pollster is underestimating the confidence in the margin when one candidate has a lead of 4% to 10%, causing it to color a state tossup when it should be lean, and lean when it should be solid.

____________________

riverrun:

This sounds more than fantastic. Thanks mightily.

I look forward to experimenting. (First pollster to fall by the wayside on my personalised charts will Zogby interactive)

____________________

Allen:

P.S., the new chart display is very nice.

____________________

DecaturMark:

This is very helpful. Although, now everyone will use this to skew the numbers to justify their particular lean. But it is a great advancement. Thanks for all the hard work!

____________________

FlyOnTneWall:

Love the new charts. Thanks!

Was wondering if I might induce you to comment on the latest post at FiveThirtyEight, detailing Seltzer's somewhat idiosyncratic approach to demographic weighting. She makes the case for adjusting her results for the under-30 vote, in particular, as a means of compensating for her inability to sample cell phone users. (The Pew results make me wonder if even this is a sufficiently aggressive move, given its suggestion that you can't simply multiply out your results among landline users to reflect the full population - those using cell-phones seem to be disproportionately supportive of Obama,)

Her second observation is even more provocative. She suspects that, in addition to whatever reverse-Bradley effect we may have witnessed earlier in the cycle, black respondents are becoming less likely to self-identify as such. That's fascinating. In cycles past, it's been white respondents who have bristled at the question, and most pollsters treat the "declined to respond" category as a virtual proxy for white/caucasian. But it makes sense - we've already had evidence that black respondents may be less likely to reveal their actual voting preferences for fear of being perceived as voting along racial lines; it stands to reason that the same effect would work in reverse, and that respondents who are revealing their support for Obama might be less likely to identify their race.

Your thoughts on these two points?

____________________

thoughtful:

@Boomshak

Now you will be able to give it off the Boomchart!

____________________

Thanks!

While I have to thank you repeatedly for these new charts, the many different ways we can run our own regressions are only going to give us tools to argue with you a lot more. The charts themselves are great, but your bravery for putting these tools in the hands of citizens is exemplary. This is internet based empowerment at its best!

Thank you, thank you!

____________________

brian d:

Can't seem to change the date range in them - keeps giving me an invalid date range when I tried to change it for both the Virginia and National charts.

____________________

Thanks! A very cool addition to the website. When you set of tools.

An addition that might be useful to add is the estimated margin-of-error or a color coding to address Allen's posting, above. If you set the sensitivity to more, you can more clearly see a short-term movement towards Obama in many states. But I can't tell if the movement changes the color classification, particularly when I change the included pollsters.

____________________

Save US:

Nice... I love it

____________________

PHGrl:

Awesome new tools!

One questions -- for anyone..
do most state polls use RBS samples? or is it a mix btw RBS & RDD??

If its a mix-- could we have RBS /RDD as a filter choice?

ALSO- possible to add cell phone inclusion as a filter- (for RDDs)...

____________________

Herb Caudill:

The interactivity is great. The mouseover features aside, though, the new chart is far less readable as a chart than the old one.

It's much harder to read the trendlines - and especially to get a sense of the gap between the two, which is the single most interesting number conveyed by this graphic. The reason is that the 3D "bubbles" create a lot more visual noise than the flat-colored circles did in the old chart.

____________________

hobetoo:

Oh how cool. Thank you for putting so many new toys on our playground!

____________________

nimh:

Brilliant, very nice. Thank you very much.

____________________

nimh:

As Erik Hare said, I particularly like the degree to which you have handed users a fair extent of power to make their own choices in how to visualise the trends. The interactivity is great.

But now that I'm looking at the actual charts, I do also have to agree with Herb Caudill. The new chart is far less readable: less crisp and clear. And this is largely due to how the 3D "bubbles" create a lot more visual noise than the old circles did. That's a really good point, and I hope you will still tweak the design a bit in that respect.

____________________

Justin:

This is great, Mark. Thank you.

____________________

Babbit:

Sweet!!

*runs off to play with new tools*

pollster.com folks, you rock!

____________________

Mark Lindeman:

@brian d: I'm able to change the date range in the national chart without any problem. Maybe your problem is browser-specific, or maybe it has something to do with the date range. Can you provide more details?

@FOTW: Not that you asked me, but methinks Selzer is doing the reasonable thing given the absence of cell phone data. It obviously isn't guaranteed to substitute for a fair cell phone sample. The key, as Nate Silver emphasizes, is that the weights are applied before the likely voter screen -- so the youth share of the electorate is based on the screen, not on prior assumptions.

@Herb and nimh: Yeah, I guess all those "bubbles" on the graphs are a bit busy. The currently highlighted poll, if any, does need to stand out, but the others needn't.

____________________

Allen:

The new display tools are very nice. After experimenting, I have to say I think the display more accurately reflects the true state of the race if you set filter to no internet polls, and set smoothing to more sensitive. I think those should be the defaults :-)

____________________

Wow, what an incredible tool. A long time lurker, I was compelled to comment.

I'm sure you are working out the kinks, but I caught a couple buggy issues.

1) the date range tool gave me a lot of trouble. Even when I finally got the format and ranges correct and got it to render one time, it stopped working. I think it may be rejecting dates if there is no data for them but I'm not sure.

2) The % range was a little confusing at first. It would be better if it automatically rounded up or down and perhaps allowed 5's as well as 10's. All in all it is a great tool and was working fine.

Thanks again -- this is going to be a lot of fun and, dare I say, useful.

____________________

mattt:

First post.
Really appreciate the effort and commentary on this site. Your drill down to state details, posting of recent polls, and top ten must reads are very helpful.
Thanks for your efforts!

____________________

serpounce:

Wow, you guys are just feeding the addiction now aren't you? I already spend too much time finding around on this site!

____________________

slinky:

Yes, java script is very nice. We've been using it for a number of applications, for a litte over a year. Thank you for your balloons.

____________________

ReprobateMind:

Very nice.

Two suggestions:

1a) Ability to set personal defaults for each of the tools.
1b) Ability to set current as defaults.

2) Auto scale y axis (%) to fill chart area (5% increments would be fine).

____________________

FlyOnTneWall:

Mark:

Thanks for the response, anyway. Perhaps the most enlightening part of the Seltzer conversation was her willingness to drop her veil of secrecy, and tell us how she does her work. I know it's a longtime crusade for (the other) Mark, and appreciate his efforts on the subject, but it's nice to see a pollster be this forthright. I really wish that she'd drop the Omertà and name names, though - if pollsters are working with inherently flawed models (and she clearly believes they are) then the public deserves to know that. It's the only way to force change in the industry. Wouldn't you just love to know how, say, Zogby constructs his samples?

Seltzer shows a huge Obama lean this cycle. I wish she'd disclose these kinds of things with each polling release, and not just in a technical aside. It'd be utterly fascinating to look inside her black box, and see to what extent the leads she's identified are artifacts of her assumptions - and then to reconfigure data from other pollsters using the same assumptions, to see if it'd produce the same sorts of leads.

At any rate, keep up the great work - I'm going to play with the new toy you guys have given us some more.

____________________

Mark Blumenthal:

All: Thanks for the compliments, suggestions and (yes, darn it) bug reports. We will try to squash them. And apologies to all for the problems with the embed function -- we've temporarily disabled it until we get it ironed out.

Brian D: Can you be more specific about the date range problems you're having?

Is anyone else having issues with the date or percent ranges? If so, could you either post a comment or email us (questions at pollster dot com), and please be as specific as possible about what you are trying to do and what isn't working?

____________________

RS:

Hey Mark B. et al.:
Great tool - amazing, really. Kudos to all involved!

I must say though, that this makes the trendlines really transparent - not necessarily to their credit! I was playing around with the CO map, and it turns out that the TargetPoint/NewLeadership (R) poll taken in March/April (!) this year (McCain 51-39) changes Obama's current numbers from 49.1% (included) to 47.6% (excluded)!

Other similar effects, e.g. live interviewer polls show Obama in the lead when including TargetPoint, but McCain leading when excluding TP! Within MOE, of course, but the trends are different!

So it looks like a single poll taken almost 6 months back can have an almost major effect on the current numbers... Weird.

But congratulations on the transparency! You folks rock!

____________________

Mark Blumenthal:

Fly: Not ignoring you, just a little behind on the analysis side today. Will get to it.

____________________

jme:

Well done, pollster!

I only have a few suggestions:

- Simply from a usability standpoint, if you enter an ending date past the most recent poll, the script should interpret that to mean "show me polls up to the most recent", rather than kick back an error.

- I second the aesthetic opinion expressed by several others that the slightly 3-d effect of the dots is a step backward (we're really picking nits, here, aren't we?). Personally, I might opt for smaller dots and thinner lines, but the issues involved in making these things look good across browsers/computers might be fairly complicated.

- I applaud the inclusion of more tick marks along the bottom by month; a notable deficiency of your old plots.

- I also second Michael McDonald's suggestion for a way to see how the lean/tossup classification changes as you fiddle. Although, based on recent conversations here, that methodology may be under review, so that might be a longer term project.

Anyway, looks good!

____________________

Mark Blumenthal:

@Herb Caudill and any one else who found the 3d effect on the points distracting:

We did a quick revision, dropping the 3D affect. What do you think? Better?

Anyone miss the 3D data points already?

____________________

RS:

I do :-(
I thought the 3D dots were sooo pretty...

____________________

jme:

RS: "So it looks like a single poll taken almost 6 months back can have an almost major effect on the current numbers... Weird."

I've commented (obliquely) on this point recently. Basically, this stems from one's methodological choice, and is, in a subtle fashion, unavoidable.

The difference between pollster and RealClear is an excellent example. Basically, pollster fits trend lines to _all_ the polls, whereas RCP simply averages the most recent polls.

Technically, the end of the trend lines on pollster are _also_ (weighted) averages of the most recent polls as well, but there's an important difference: pollster's trend lines are trying to match up to _all_ the data at the same time. So imagine the cloud of points representing the most recent polls. Polls from a long time ago can influence how the trend line flies into this cloud of points.

The end result is that it will be harder for a trend line to "forget" old data, compared to RCP's method, which literally drops polls after a set period.

There's no clear cut answer here for what's best. It all comes down to what's called the "bias - variance tradeoff". (Apologies if this is all old hat; I have no idea what your stats background is)

Essentially, our error (mean squared error, MSE) of the trend line at any point can be broken into two pieces. It's:

MSE = Bias^2 + Variance

In general, efforts to reduce one will increase the other. Hence, pollster's trend line estimates might exhibit more bias (i.e. they will be slower to reflect "reality" at the endpoints) than RCP; but on the other hand will also likely exhibit less variance (they won't flop around quite as much in response to outliers).

An excellent (post election) topic for Charles (or someone with some time on their hands) would be to work backward through pollster's trend line estimates and look at the difference. Specifically, you'd walk backward through time from election day and look at (1) pollster's trend line estimate _up to_ that day, (2) a RCP simple average of, say, the 8 most recent polls up to that day and (3) the "true" estimate based on a trend line fit to the full set of polls through election day.

My prediction would be that compared to (3), (1) would go through periods where it was "biased", i.e. slow to react, and (2) would have days where it jumped around needlessly (high variance). But then, I'm speculating here based on theory.

Further, I suspect that if pollster's method is at a disadvantage at all, it will be in situations where there are a limited number of polls. With only a few polls, there isn't enough time for the trend line to "reset" before entering the cloud of the most recent polls. Increasing the trend line sensitivity can help here, but only up to a point, probably.

Anyway, that was probably more info that you wanted...

____________________

Mark Blumenthal:

All: By popular demand, we've already flattened out those 3D data points. What do you think?

We should also have fixes to the various axis problems within a few hours.

[Note: You can tell I need some sleep...sorry for the semi-double post!]

____________________

jme:

Dots are much better now...still would prefer thinner lines though...

____________________

RS:

@jme:
That's an interesting explanation - thanks. I have some, but still limited background in statistics, FYI.

I don't know if you looked at the CO Pres-poll page, but your comments and the TargetPoint poll I was talking about points me to one possibility. The TP poll was taken when there were fewer polls; recently, there has been a wave of polling in CO (relatively speaking).

So I wonder if extra weight is placed to this poll, as an outlier - the biggest (and only) McCain lead (+12) among polls of that "few polls" vintage - Ras had the race tied or O+3 on either end of TP, while the others generally had Obama ahead.

Making the trendlines more sensitive, as you suggest, gives (IMO) a more realistic picture - Obama +4 irrespective of whether TP is included or not. I suppose "more sensitive" means more sensitive to recent polls, as margin-wise TP has little effect with high sensitivity (unlike in the "default sensitivity" case, where Obama's lead drops from 3.2% to 1.7%).

Fun times... damn you, Pollster.com! (for giving me more things to play with and procrastinate ;-))
[Said he in his best Charlton Heston imitation...]

____________________

jme:

@RS

As a further note, as far as I can tell (Charles has been fairly transparent, but I could have misunderstood), these trend lines use a single smoothing parameter for the entire trend line.

In principle, it is possible to fit trend lines that adaptively select the degree of smoothing locally.

I think Charles mentioned in another thread that the choice of smoothing parameter is more art than science (than he would like); ideally we'd like something the doesn't require us to "fiddle with the knobs" to get it just right. There's always the danger that I really like a particular setting because, you know, I'm a huge Obama supporter. Sadly, that's just human nature.

In any case, adaptive smoothing procedures are possible, but would probably necessitate _more_ fiddling to get it "just right". And the "just right" setting would probably be even more different between states and between different races.

And of course, one might implement some complicated cross-validation procedure to automatically select the adaptive smoothing parameter...

And on and on and on...

____________________

jme:

@RS

Don't mean to be responsible for like half the posts here, but...

Check out MN for another wacky example of the phenomenon you noticed. Specifically, the removal of a single Zogby Interaction poll from _June_ increases Obama's lead from 3.2 to 4.8! (Using the standard smoothing setting)

And in this case you can clearly see that this one Zogby poll pulled McCain way down in early summer. Hence his trend line was _already_ predisposed to trend upwards. Basically, it changes the slope at which McCain's trend line hits the most recent polls. The steeper the upward pitch you have going when you enter that cloud, the more strongly his trend line is predisposed to react to those polls.

I wonder how many cool things like this pollster.com will find just from commenters pouring over this new toy?

____________________

RS:

@jme:
I am an Obamabot with the best (worst?) of them... :-)
My figleaf: the complaint was more along the lines of "a single poll should not influence the final results much."

The Zogby MN poll you point out is also the only one with Barr at much higher than zero - see how Barr drops to zero by excluding Zogby Internet? For all we know, that may be the *only* realistic poll for that very reason!
[MPR/Humphrey has Barr at 1%, Nader at 3%.]

Still, I wonder if Pollster.com should throw out pre-June polls... Just a thought. Oh wait - or not. I thought the "date range" changed the period over which the polls were "averaged" - apparently not. That's just a zoom feature... oh well!

See, every time Mark B and the good people at Pollster.com put something out, all I want is even more!!

____________________

James:

Now that the charts are interactive, though, couldn't you give users the option of switching between different methodologies and marveling at all the differences that occur?

Also, I think the old charts were clearer, more readable. The flat circles as data points are better, but I also liked the brighter, more contrasted colors of the previous charts. And I think the lines shouldn't be quite as thin as before, but a little thinner would make for a clearer chart, I think. When you combine darker colors with thicker lines, the charts don't look as alive and active as before, which is quite ironic.

And on customized embedded charts, please make sure it's clear that a chart has been customized, ideally making it easy to switch back to the default view in one click. You don't want people to be able to use Pollster's name for misleading purposes.

Those details aside, though, this is absolutely fantastic. I remember, even before Pollster came along, thinking someone should create an interactive poll charting system where you can toggle different pollsters and make other adjustments. I'm glad someone finally did it, and I'm glad that someone is you! All hail the new clear leader in online polling data!

My wish now is that after the election you expand into more topics (like pollingreport.com does). You'd have to give Eric a raise, though!

____________________

ReprobateMind:

Yes, I too have problem with the date range tool.

The default date range is 01/01/2008 – 11/15/2008. Even if I just click the "Set Range" button without changing the default dates, it fails with an "Invalid Range" message.

Also, not able to move cursor between month/day/year fields (have to mouse click to move between the fields).

____________________

jme:

Wow, I just keep noticing the same thing...in VA this time (to a _much_ smaller degree), but exclusion of a very old Zogby Internet poll that was bad for _McCain_ reverses the order of the trend lines at the current estimate. (The difference here is tiny, but it's just very instructive to see how old results in times of sparse polling can influence the trend lines!)

____________________

Herb Caudill:

Wow, that was quick. Thanks for getting rid of the 3D dots. Tufte would approve.

____________________

Thanks to all for the enthusiastic vetting of the new charts and the informative discussion of the effects of changes to the data.

I expect you to find many instances where changes even to past polls make some difference to the current estimate, though that should be especially apparent where the number of polls is more modest and the poll being excluded is an outlier or at least relatively extreme.

We could argue endlessly about which is the "best" representation of the data... come to think of it, that is sort of the point in letting you change the charts to your heart's content. Let a thousand analyses bloom!

Charles

____________________

jme writes (sorry for the long quote but I want to comment on multiple parts:)

"As a further note, as far as I can tell (Charles has been fairly transparent, but I could have misunderstood), these trend lines use a single smoothing parameter for the entire trend line."


chf: That's right, it is a global smoothing parameter that adapts to the number of polls in a state (so one state might smooth differently than one with either more or less polls) but the smooth is constant over time within state.


"In principle, it is possible to fit trend lines that adaptively select the degree of smoothing locally."


chf: That's an interesting idea and worth future development work. The data are obviously "clumpy" and generally denser later in the campaign.


"I think Charles mentioned in another thread that the choice of smoothing parameter is more art than science (than he would like); ideally we'd like something the doesn't require us to "fiddle with the knobs" to get it just right."


chf: The problem is that objective fit measures develop a "flat" region across a disturbingly wide range of smoothing parameters. The optimization problem has a variety of nearly equivalent measures of fit which nevertheless produce quite different visual fits. See Loader's"Local Regression and Likelihood" book for one treatment of this. The conclusion there is that visual inspection is better than an algorithm once you are within the flat region of the objective function. That has been my experience as well over a wide range of variables and series lengths.


"There's always the danger that I really like a particular setting because, you know, I'm a huge Obama supporter. Sadly, that's just human nature.

In any case, adaptive smoothing procedures are possible, but would probably necessitate _more_ fiddling to get it "just right". And the "just right" setting would probably be even more different between states and between different races."


chf: Right again. One reason I have a function compute the degree of smoothing is to avoid subjective judgements, especially after seeing the fit with different amounts of smoothing and picking the one I "like" better. But that means the missing out when a different degree of smoothing would "look better".


"And of course, one might implement some complicated cross-validation procedure to automatically select the adaptive smoothing parameter..."

chf: Been there. I wrote an optimization routine giving a choice of cross validation, AIC or BIC as fitting criteria. Alas, Loader was proven right-- these produced visually unacceptable results ("obviously" under smoothing sometimes and over smoothing others) and the three criteria often couldn't agree and resulted in considerably different recommendations for smoothing even of the same data.

____________________

nimh:

Much appreciated for getting rid of the 3D bubbles for individual data points. Looks better.

I do think the new charts still look less crisp than the old ones, but I cant quite put my finger on why anymore - the most obvious problem has been removed now anyway!

____________________

Mark Blumenthal:

An update:

1) As of this hour, as noted in the top post, the embed bug should be history and the tool should be working properly.

2) We did find a bug that was causing the date range problems mentioned above. That should be eliminated now.

3) While not a bug, we confused everyone by setting the percentage increments to 10%. You can now set the percentage range in increments of 5% (but not less).

If you were experiencing problems before, you might want to clear your browser cache before trying those problematic tools again (instructions on doing that here).

If anyone continues to experience bugs or other problems with the charts, please do not hesitate to post a comment or email us at questions at pollster dot com.

____________________

jme:

Thanks for the lengthy response, Charles. I've had similar problems (with other types of data) using CV to choose a local smoothing parameter.

I wonder if it would be better behaved if you gave it fewer choices of smoothing parameters. Namely, have it only choose, locally, between the three options (low,med,high) in your new polls. Or you could create an ad hoc criteria for switching between the three based on the poll density (clumpy-ness).

____________________

RS:

I just noticed that by clicking on one poll, e.g. SUSA, ALL of SUSA's polls are linked by dashed lines, showing the trends from that pollster.

WOW! THIS IS AWESOME!

[Sorry, but I just had to yell!]
p.s. Any chance we can get the user to decide the date range over which the trendlines are fitted? :-)

____________________

James:

I'm liking the look of the chart better today. Did something change, or am I imagining things?

Also, count me as another vote for dynamically adjusting the y-axis to fit the data points, rather than having large white spaces above and below. I think it looks more exciting that way.

____________________

Herb Caudill:

A couple more suggestions in case anyone is still checking this thread:

1. I'd like to see a separate line chart along the bottom showing movement in the gap between the two candidates, like RCP does:
http://www.realclearpolitics.com/epolls/2008/president/us/general_election_mccain_vs_obama-225.html

2. Could we show a little variation in the diameter of the discs to reflect the relative weight of each individual poll? There's obviously not room to do much - probably just between 2 and 6 pixels - but it would be an efficient way of adding more information to the chart.

____________________

jceaser:

I don't get it. Why can't I see the chart. All I get is a list of polls. The classic chart works but nothing else. This is why I hate flash.

____________________

Farleftandproud:

Rasmussen are such fucking liars. I think their polling on party ID is about as unbelievable as you can get. They are telling people that 32 percent of Americans identify themselves with the GOP. They probably picked all the states that Mccain won, and didn't include, VT, Mass, NJ, and Ca. I think the grass could be blue and Rasmussen would tell you it was green. If they were to do a weather report, they could say the forecast was 80 degrees and beautiful and the weather ends up being 63 and rain. It is unbelievable just how stupid they think the American people are.

____________________

I have been a regular reader of Pollster.com for several months. It has now turned to crap! Is it just because you are not done installing it? Please return it to the easy-to-use format that it was before!

Egor

____________________



Post a comment




Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR