FlowingData Forums » Tag: graphs

FlowingData Forums » Tag: graphs - Recent Posts http://forums.flowingdata.com/ Strength in Numbers en Sat, 11 Feb 2012 18:47:27 +0000 Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2184 Wed, 20 Apr 2011 21:04:45 +0000 caredwen 2184@http://forums.flowingdata.com/ Tyson - I see your point about "lineyness". One of the challenges is ensuring that we appropriately emphasize the broad comparative group. The way I see it, the two most important pieces of information on the chart are the funder's current average and the overall median, so I want to make it as easy as possible for a viewer to make that comparison. You're right that it gives one a lot to look at, though. Do you have suggestions for ways to make this comparison pop without cluttering the field too much? Perhaps just a bigger box plot? :-) On the same subject -- I like box plots, but for the smaller cohort they are challenging, because those cohorts tend to be relatively small (comprising between 12 and 20 funders). Showing the interquartile range of a small group like that may be too much information relative to just showing the max, min, and median - one set of quartiles may be enough for a viewer to contend with... Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2182 Tue, 19 Apr 2011 14:06:31 +0000 tyson 2182@http://forums.flowingdata.com/ Your new chart is clearer than the first, but I still think it's confusing. First off it's a little too "liney." All the lines compete with each other and it's hard to pick out what's important. Let me illustrate: <a href="http://www.flickr.com/photos/63648197@N00/5635541664/" title="5632037355_a34a8a8a36_b by fotoman607, on Flickr" rel="nofollow"><img src="http://farm6.static.flickr.com/5142/5635541664_06349e14bd.jpg" width="481" height="500" alt="5632037355_a34a8a8a36_b" /></a> When you simplify your chart down to shapes you should be able to pick out what you want to be most important and relevant to your viewers. When looking at yours nothing seems to really stand out from the rest of the chart which makes it hard for your viewers to figure out what they should be looking at. <a href="http://www.flickr.com/photos/63648197@N00/5635539524/" title="funders_thresh by fotoman607, on Flickr" rel="nofollow"><img src="http://farm6.static.flickr.com/5109/5635539524_1e676c2485.jpg" width="500" height="400" alt="funders_thresh" /></a> This one is much clearer because at first glance you can immediately tell what shapes you should be looking at. Second, is there a reason against using box plots? It seems to me that if you are trying to show a median, range, and middle 50% the box plot would be the best way to go. A box plot would also show outliers, bringing more information into the chart. I can see arguments against using them because they might be above the level of understanding of your viewers (I work for a newspaper and probably wouldn't be able to use them). On the other hand you are already basically showing everything a box plot shows but by drawing all those line across the chart you really busy things up. I guess you have you ask yourself if you can count on your viewers to read between the lines (literally) and do without the lines running all the way across or keep the lines in there and risk having more viewers not understand the chart at all. Also, as Thorri also did in his chart I think that since your data for "foundation X" is time based it should be shown in sequence connected with a line. Stacking them on top of each other, although it takes up less space completely removes the ability of your viewers to see the change of the data over time. However, the dots that show the programs can be left stacked up because they do not have a sequence. Finally, Joshua brings up a good point in that instead of rethinking how to redraw the chart you already have to be a little more understandable, it might be more helpful to start from scratch. Go back to your core data and pick out what exactly you want the chart to show and there might be a better way to do it. Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2181 Mon, 18 Apr 2011 18:53:18 +0000 Joshua de Haseth 2181@http://forums.flowingdata.com/ Hi, I don't know precisely what data you got, but if you got data by time I would try to use the data in the context of time. As we all know one year of data (or one data point in a year) is pretty dangerous to work with. Ok you can compare the performance of one fund to all the others and a subgroup, but only for one year. What if something remarkable happened to all the others except your fund, if you see the results you will go crazy and will try to do anything to get back in line with the others. But if you see that years results in comparison to the other years it seems that something out of the ordinary has happened. This also could mean you have to take action, but your reaction is of a totally different kind (I hope I'm clear enough). This is just an example ofcourse. I just want to make a point that trend data is probably a better way to go, if available. In a line diagram you can fit all the info in you already got. You can still compare your fund to the range of all funders (grey area), the middle 50% (between dashed lines), the range of cohort funders (vertical lines at each year), the median and the three different programs. And all this for all the years next to each other, so much more information. To make this example clearer you could experiment with the colours and labels ofcourse. At first it looks like a lot of lines, but reading the labels connected to the lines this should clear up pretty fast. <a href="http://www.flickr.com/photos/61891160@N07/5633130522/" rel="nofollow">http://www.flickr.com/photos/61891160@N07/5633130522/</a> Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2180 Mon, 18 Apr 2011 15:30:44 +0000 caredwen 2180@http://forums.flowingdata.com/ I'm the research analyst Kevin mentioned in his above post, and Tyson, I thought your suggestion of segmenting the data more horizontally was really helpful. I adapted it a little, below. The modifications I made in this design served two purposes: 1.) to make the current average rating the most salient piece of information on the chart 2.) to play up the overall comparative cohort more, relative to the smaller comparative cohort <a href="http://www.flickr.com/photos/61912578@N04/5632037355/" title="new+chart+1 (1) by caredwen, on Flickr" rel="nofollow"><img src="http://farm6.static.flickr.com/5267/5632037355_a34a8a8a36.jpg" width="481" height="500" alt="new+chart+1 (1)" /></a> Thoughts, critiques, other suggestions? p.s. I'd also like to second Kevin's thanks to the community - it's been a pleasure reading through these thoughtful contributions! Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2179 Mon, 18 Apr 2011 14:24:16 +0000 kmbolduc 2179@http://forums.flowingdata.com/ Wow, folks, thanks so much for the feedback. Every post has been immensely helpful. I am so excited by all these sketches and mock-ups. I'm also realizing that my goal to condense everything into less space is taking away from the understandability of the charts. I love the granularity of the normal curve. And tyson's suggestion has gotten me thinking about what info we could also just put in a tiny table next to the primary info in the graph. (One of our research analysts was so enthused by Tyson's concept that she's playing around with it now. I'll have her post her take on this as well.) The next step is for us to present some mock-ups to our advisory board in the beginning of May. I'll continue to monitor this post in case others have further suggestions, and I'll post the mock-ups we prepare for the advisory board. This blog and its community is fantastic. Thank you! Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2175 Fri, 15 Apr 2011 06:15:06 +0000 Thorri 2175@http://forums.flowingdata.com/ Yup, I think Tyson nailed it. I'd go with that. Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2174 Fri, 15 Apr 2011 03:00:56 +0000 tyson 2174@http://forums.flowingdata.com/ I feel like your chart is so compressed you can't make any meaningful comparisons between the data because you get lost. Here is my quick attempt at cleaning things up a little: <a href="http://www.flickr.com/photos/63648197@N00/5620837485/" title="funders by fotoman607, on Flickr" rel="nofollow"><img src="http://farm6.static.flickr.com/5183/5620837485_6a0da5b962.jpg" width="500" height="400" alt="funders" /></a> - I used boxplots to show both sets of your funders. A little different than how you were doing it but boxplots are something that more people have seen before so it's easier to understand the chart at first glance. It's also much cleaner in my opinion. - Because you want to make comparisons between all the data I kept everything in one chart using a common axis. Because you had separate categories of data I split the one chart into separate panes. This allows you to make comparisons between the data, but not get lost in it. - My chart could definitely be squished up a bit horizontally if you needed to make it smaller (might even look better that way). Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2173 Thu, 14 Apr 2011 22:03:41 +0000 nathany 2173@http://forums.flowingdata.com/ Wow actual sketches and mockups. You guys are awesome. Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2171 Thu, 14 Apr 2011 18:05:32 +0000 Thorri 2171@http://forums.flowingdata.com/ I think it would help not to try to put all of this into one type of chart. Also within one chart, to seperate the data to avoid confusion. Below is a link to my sketch. I think the data may even be seperated further. (I'm thinking of the Arts/Education/Health points). Hope this helps :) <a href="http://dl.dropbox.com/u/65944/CEP.png" rel="nofollow">http://dl.dropbox.com/u/65944/CEP.png</a> Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2170 Thu, 14 Apr 2011 15:40:36 +0000 sacoope 2170@http://forums.flowingdata.com/ I think the first step is to try and get rid of the legend and incorporate it in to the visual. The shapes and colors aren't a bad idea, but the viewer has to reference back and forth between the chart and the legend - which may ultimately lead to confusion. Have you thought about using a labeled distribution curve? Depending on the audience this may look a bit too technical, but it provides an easy way to label in a more user friendly manner. I've attached a link to a quick sketch-up I did (based on a normal distribution - your data may not look like this). Please ignore the lack of formatting and coloring - just trying to give a concept :-) Hope this helps! <a href="http://www.flickr.com/photos/61747317@N05/5620242624/" rel="nofollow">http://www.flickr.com/photos/61747317@N05/5620242624/</a> Nonprofit Looking to Improve Data Charts http://forums.flowingdata.com/topic/nonprofit-looking-to-improve-data-charts#post-2168 Thu, 14 Apr 2011 13:10:07 +0000 kmbolduc 2168@http://forums.flowingdata.com/ At the heart of our work at the <a>Center for Effective Philanthropy</a> (CEP) is the provision of comparative data that allows one funder to understand how aspects of its own performance compare to the other funders. In much of our work, the manifestation of that activity is charts in our reports that display the comparative data. I’d like to ask your advice in making CEP’s data display stronger. The Center for Effective Philanthropy’s mission is to provide data and create insight so philanthropic funders can better define, assess, and improve their effectiveness – and, as a result, their intended impact. We’re a nonprofit organization based in Cambridge, working many of the largest philanthropic funders in the country. Here are two examples of the way our charts look now. <a href="http://www.effectivephilanthropy.org/assets/blog/chart1.2.png" rel="nofollow"><img src="http://www.effectivephilanthropy.org/assets/blog/chart1blog.png" alt="" width="480" height="290" /></a> <a href="http://www.effectivephilanthropy.org/assets/blog/chart2.2.jpg" rel="nofollow"><img src="http://www.effectivephilanthropy.org/assets/blog/chart2blog.png" alt="" width="480" height="290" /></a> We’ve tweaked these charts here and there over the past few years, but we still hear from some funders, and in our 3rd party feedback, that these charts can be tough to understand. On the other hand, some of the funders we work with love these for the amount of information they pack into a small space. We think we can improve these charts. Although any change may not be immediate, we want to brainstorm now some other possibilities. Fundamentally, any data display we use has to meet just a few basic parameters. It must: • display comparative context so that one funder can consider its relative results compared to our database of others’ results • simultaneously display both an absolute scale and relative results (because both are necessary pieces of information in interpreting results) • be flexible enough to display a potential segmentation of the overall data, display trend data, and (probably) also a sub-group comparison – a cohort from among the full dataset This blog is one that we follow and often discuss as a staff at CEP. We’re huge fans of interesting data visualization, and we admire the ideas and examples posted in this blog. It hit us recently that you all might have some incredible ideas about our work. We’d welcome the advice. Or if the community has advice about a great data visualization expert have worked with, that would be fantastic too. (This is too small a project for the couple we’ve reached out to.) What’s in it for you? The knowledge that you aided CEP in its efforts to help foundations become more effective. And, if we choose an idea you suggested, a very public and grateful acknowledgement for your efforts and ideas. Thanks in advance for your help. Quick reference for graph and dashboard design- print, Android and iPhone http://forums.flowingdata.com/topic/quick-reference-for-graph-and-dashboard-design-print-android-and-iphone#post-2153 Tue, 05 Apr 2011 11:47:17 +0000 stephen.mcdaniel 2153@http://forums.flowingdata.com/ Freakalytics has published to Amazon, Android and iPhone a new quick reference for building better graphs and dashboards. This app is complimentary on the Android and iPhone thanks to our sponsor, Tableau Software. Note that the Android is now released and the iPhone is several days away from release. The Amazon version, “Rapid Dashboards Reference Card”, has all 64 tips readily available on four full-color, laminated pages. The card serves as a handy reference for yourself, your team and even your business audience during design meetings. Read more at <a href="http://www.freakalytics.com/books/" rel="nofollow">http://www.freakalytics.com/books/</a> Best regards, Stephen McDaniel Co-founder, Freakalytics Faculty, American Marketing Association Tick Marks... http://forums.flowingdata.com/topic/tick-marks#post-1954 Mon, 18 Oct 2010 09:14:24 +0000 dan_c 1954@http://forums.flowingdata.com/ There are a couple of related issues that I have encountered over the years. If the data are thought of as perfectly discrete, as if it were irrelevant that the x axis is a time dimension, then it seems straightforward to have a one-to-one correspondence between ticks, labels, and data points, with ticks and labels possibly suppressed for the sake of clarity. In that case connecting the dots with lines is arbitrary and unnecessary. You certainly wouldn't do that if the x axis represented the 50 states, say. If there's one tick mark and one label for each data point (with ticks and labels possibly suppressed for clarity), then there doesn't seem to be any significance to the regions between the observations, and no reason to connect the dots other than "convention." Skinny bars would seem to be more accurate and would exclude no information. But for time-series charts, you can conceive of the x axis as a continuum, with ticks chosen to identify points in time. The definition of "point" depends upon the data, but in this case you might consider a point to be a particular date (even when plotting annual data, say). For some financial charts, a point in time could be a millisecond. You can think of the Y axis as a measurement of a concept that varies continuously over time but is observed only discretely. In this conception, connecting the dots does convey a specific meaning: it provides the best available representation of the variation over time. How this happens depends upon how the data are observed. You might have data that represent the average or sum of a period (month, year), or the beginning or end. (These are the most common, but there are plenty of other possibilities.) In this formulation the regions between the ticks do have significance, as intervals, the points in time between those chosen for ticks. So these labels can sensibly go between the tick marks. So in this model, although the ticks refer to (selected) points in time out of a continuum, the labels can refer to intervals (such as months, years, quarters) that collectively compose the entire range of time. Observations can be plotted to correspond with intervals or with data points: if the (technical interpretation of the) observation corresponds to the whole month or year, say as an average or sum, then it makes sense to plot the data above the label, between the tick marks identifying the beginning and end of the period. If the observation by definition corresponds to the end of each period, then it makes sense to plot the data over the tick mark corresponding to the end of the period. But in either event, there is the choice to label the interval (such as 2007) between the ticks, or to label the tick itself (as in Dec 31, 2007). This approach seems to be the best way to handle joint plotting of data that are observed in different frequencies. Sorry for the length. I hope this adds some value (or that somebody finds this late contribution). Tick Marks... http://forums.flowingdata.com/topic/tick-marks#post-1888 Sun, 19 Sep 2010 23:36:30 +0000 nathany 1888@http://forums.flowingdata.com/ yeah... that forums are kind of flawed. haven't had the time to make it better. Tick Marks... http://forums.flowingdata.com/topic/tick-marks#post-1885 Fri, 17 Sep 2010 06:51:32 +0000 ryanhbowman 1885@http://forums.flowingdata.com/ Thank guys, especially on the tick labelling. Sorry to be so late, I had assumed the site would email me. Should have checked back...but maybe a feature for the next version Nathan? Or maybe I neglected a option? Thanks again... Tick Marks... http://forums.flowingdata.com/topic/tick-marks#post-1875 Wed, 01 Sep 2010 13:28:33 +0000 DundasData 1875@http://forums.flowingdata.com/ Most often tick marks are placed outside of the axis line, but either way is acceptable. Good point about using tick mark labeling Nathany. I personally find it irritating when I have to figure out what axis values tick marks actually represent. Tick Marks... http://forums.flowingdata.com/topic/tick-marks#post-1865 Wed, 25 Aug 2010 12:02:07 +0000 nathany 1865@http://forums.flowingdata.com/ Good questions. I think it's more a matter of taste with this one. Personally, I do (b), but label the tick marks themselves rather than the space in between them. I feel like when the marks are inside, it almost seems separate from the labels. Tick Marks... http://forums.flowingdata.com/topic/tick-marks#post-1863 Wed, 25 Aug 2010 07:39:09 +0000 ryanhbowman 1863@http://forums.flowingdata.com/ This is pretty specific and may be so pedantic as to be pointless but I am wondering about the perfect placement of tick marks on a simple line or bar chart. Should they push into the graph (a) or push out into the axis label (b)? <a href="http://www.flickr.com/photos/71494118@N00/4925883981/" title="Tick Marks on Charts by ryanhbowman, on Flickr" rel="nofollow"><img src="http://farm5.static.flickr.com/4096/4925883981_bb6e13f68d.jpg" width="500" height="229" alt="Tick Marks on Charts" /></a> List of graphing packages for websites? http://forums.flowingdata.com/topic/list-of-graphing-packages-for-websites#post-1096 Wed, 26 Aug 2009 18:55:34 +0000 PrplHaz4 1096@http://forums.flowingdata.com/ Adding some more "classic" commercial packages to the mix: ChartFX (<a href="http://www.softwarefx.com/" rel="nofollow">http://www.softwarefx.com/</a>) Charting.Net (<a href="http://dotnetcharting.com/" rel="nofollow">http://dotnetcharting.com/</a>) DundasChart (<a href="http://www.dundas.com/" rel="nofollow">http://www.dundas.com/</a>) large network layout and visualization http://forums.flowingdata.com/topic/large-network-layout-and-visualization#post-814 Wed, 08 Jul 2009 09:26:16 +0000 seb 814@http://forums.flowingdata.com/ Hi, Gephi meets exactly your needs, and the video tutorials makes it really easy to use : <a href="http://gephi.org/support/tutorials/" rel="nofollow">http://gephi.org/support/tutorials/</a> About supported file formats: <a href="http://gephi.org/support/supported-graph-formats/" rel="nofollow">http://gephi.org/support/supported-graph-formats/</a> <a href="http://www.flickr.com/photos/gephi/3176867897/" title="Gephi Screenshot C.Elegans de gephi_org, sur Flickr" rel="nofollow"><img src="http://farm4.static.flickr.com/3111/3176867897_1b7b6feeb1.jpg" width="500" height="299" alt="Gephi Screenshot C.Elegans" /></a> cheers!