One of the goals I set for myself this summer was to learn a bit about D3, a visualization toolkit that can be used to manipulate and display data on the web. Considering that the trees are bare and we’ve already had our first frost here in Wisconsin, you can safely assume that I am behind schedule. Nevertheless, I feel that I’ve finally reached a point where I have something to publish, so here goes.

First of all, a little background. D3 is a JavaScript library that allows you to bind data to any of the elements (text, lines and shapes) you might normally find on a web page.  These objects can be stylized using CSS and animated using simple dynamic functions. These features make D3 a perfect tool for creating interactive charts and graphs without having to depend on third party programs like Google Charts, Many Eyes or Tableau.

I wanted to start out with something simple so I elected to go with a basic line chart using data I pulled from Pro-Fooball-Reference.com. This site contains a ton of great information and statistics from the past 90+ years of the National Football League but — for now — I just looked at the final scores of all the games played from 1920 to 2011. My first D3-powered chart is below. It shows the average combined scores of winning and losing teams for each year of the NFL’s existence.

Although this chart looks pretty simple, every element — including titles, subtitles, axes, labels, grids and data lines — has been created manually using the D3 code. The payoff is pretty nice. All of the elements can be reused and you have tremendous control over what is shown onscreen. To demonstrate some of these cababilities, I’ve added interactive overlays that show a few of the major eras in NFL football (derived from work of David Neft and this discussion thread). If you move your mouse over the graph, you will see these different eras highlighted:

Early NFL (1920-1933) – The formation of the American Professional Football Association (APFA) in 1920 marked the official start of what was to become the National Football League. This era was marked by rapid formation (and dissolution) of small town franchises, vast differences in team capabilities and a focus on a relatively low-scoring running game. At this time, the pass was considered more of an emergency option than a reliable standard. The rapid growth in popularity of the NFL during this era culminated with the introduction of a championship game in 1932.

Introduction of the Forward Pass (1933-1945) – The NFL discontinued the use of collegiate football rules in 1933 and began to develop its own set of rules designed around a faster-paced, higher-scoring style of play. These innovations included the legalization of the forward pass from anywhere behind the line of scrimmage – a change that is often called the  ”Bronko Nagurski Rule” after his controversial touchdown in the 1932 NFL Playoff Game.

Post-War Era (1945-1959) – The end of WWII saw the expansion of the NFL beyond its East Coast and Midwestern roots with the move of the Cleveland Rams to Los Angeles — the first big-league sports franchise on the West Coast. This period also saw the end of racial segregation (enacted in the 30s) and the start of nationally televised games.

Introduction of the AFL (1959-1966) – Professional football’s surge in popularity led to the formation of a rival organization — the American Football League — in 1960. The growth of the flashy AFL was balanced by a more conservative style of play in the NFL. This style was epitomized by coach Vince Lombardi and the Green Bay Packers, who would win five championships in the 1960s. In 1966, the two leagues agreed to merge as of the 1970 season.

Dead Ball Era (1966-1977) – Driven in part by stringent restrictions on the offensive line, this period is marked by low scores and tough defensive play. Teams that thrived in this environment include some of the most famous defenses in modern NFL history: Pittsburgh’s Steel Curtain, Dallas’ Doomsday Defense, Minnesota’s Purple People Eaters and the Rams’ Fearsome Foursome.

Live Ball Era (1978-present) – Frustrated by the decreasing ability of offenses to score points in 70s, the NFL began to add rules and make other changes to the structure of the game in an attempt to boost scoring. The most famous of these initiatives was the so-called “Mel Blount Rule” (introduced in 1978), which severely restricted the defense’s ability to interfere with passing routes. With the subsequent introduction of the West Coast Offense in 1979 — an offense based on precise, short passes – this period became marked by a major focus on the passing game.

Having created this first chart, I decided to build a second chart based on the ratio of average winning scores to average losing scores to see if there were any patterns.

The chart above shows how — after a period of incredibly lopsided victories — the average scoring differential settled in to a very steady pattern by the late 1940s and stayed at that level (roughly 2:1) for the next 30 years. Despite many changes in rules, coaching techniques, technology and other factors, only the pass interference rules of the late 1970s seemed to have any signifcant effect on this ratio, shifting it to just under 1.8:1 for the next 30 years.

While I had the data available, I also decided to look at the differences in average scores between home teams and away teams. The chart below plots this data along with the same overlay I used in the first chart.

A look at the ratio of average home team scores to average away team scores follows:

What’s fascinating about this chart is how quickly a form of parity was acheived among all the NFL teams. By the mid-30s, a measurable home field advantage can be seen at roughly 15%, a rate that has remained essential constant for over 70 years. Factors for this boost could include the psychological support of fans, familiar weather conditions, unique features of local facilities, lack of travel fatigue, referee bias and/or increased levels of motivation in home town players.

Thanks to Charles Martin Reid for his solution to getting D3 and WordPress to play nice.

My wife works as a speech and language pathologist at a local school district and her caseload is a mix of kids from ages 3 to 21. One of her most difficult cases involves a student with severe cerebral palsy who transferred into the district when he was 12-years-old. This student cannot speak or use his body to convey information and currently expresses himself primarily through eye contact and facial expressions. Because of this extremely limited range of abilities, his cognitive functioning is unknown.

The only real communication options involve interpretation of the student’s eye movements. Professionals can do this using a contraption called an eye gaze board, which is a simple frame to which you attach pictures or symbols. The professional sits face-to-face with the student, holds up the frame, and then prompts the student with a question. By observing where the student looks, the professional can make assumptions about their intended responses.

Needless to say, this approach has some limitations, particularly in this case. The student’s eye movements are hard for even professionals to interpret and the student himself tires very quickly. After careful consideration, my wife elected to see if the student could use an eye-tracking device in combination with a communication board or speech generating device (SGD) – a specialized computer that allows the user to build messages and relay them to others through a synthesized voice. (Dr. Stephen Hawking is a famous user of such a device.)

Users can access these devices directly using a keyboard or touch screen or they can manipulate them indirectly with a joystick, adapted mouse, optical pointer, eye tracking device, or other type of controller. The specific access method depends entirely on the abilities of the user and, in this case, there are not a lot of options. The student is quadriplegic and does not even have enough control over his head and neck movements to use switch access scanning, in which an indicator such as a cursor steps through selections automatically and the user hits a switch when the indicator lands on the desired choice. Blinking is also out for similar reasons.

A comparison of the two options shows some obvious advantages for the eye tracking option. Unfortunately, these devices are not cheap. While an eye gaze board can be assembled from five bucks’ worth of spare parts, a communication board and eye-tracking device cost about $8,000 apiece. No school district is going to spring for such a purchase these days so it became necessary for my wife to apply for a loaner and see if she could build a case for Medicaid.

.

Comparative Evaluation (Eye Gaze Board & Eye Tracking Device)

	Eye Gaze Board	Eye Tracking Device
Ease of Set-Up	Easy	Difficult
Ease of Listener Comprehension	Difficult	Easy
Y/N Response Accuracy	20-30%	80%
Number of Communication Functions	4	14
Size of Picture Field	4 pictures	12 pictures
Length of Time Before Fatigue	10 minutes	30-40 minutes
Maximum Length of Utterance	1	4+
Able to Fine-Tune Dwell Times	No	Yes
Able to Independently Introduce a Topic	No	Yes
Able to Communicate with Multiple Listeners	No	Yes
Able to Call for Attention	No	Yes
Able to Communicate with Non-Professionals	No	Yes
Able to Repair Communication Breakdown	No	Yes

.

1st Trial

The loaner — a Dynavox Vmax/EyeMax system — arrived in the last few weeks of the 2011 school year and came with some standard navigation screens or “boards” that are based on vocabulary and language ability levels. The user categories include — in order of ability — emergent communicators, context-dependent communicators, and independent communicators.

The primary choice for this case was between context-dependent, which means that the student’s ability to communicate depends on the environment, topic, or communication partner, and independent, which means that they are able to combine single words, spelling, and phrases together to create novel messages about a variety of subjects.

.

Examples of the Context-Dependent “Child 12″ Navigation Page (left) and Scene (right)

.

Examples of the Independent Gateway “Child 12″ Set-Up (left) and “Child 40″ Set-Up (right)

.

These navigation boards make extensive use of picture communication symbols (PCS) and the Fitzgerald color coding system for language development. PCS are simply standard graphics whose meanings are easily understood while the Fitzgerald “key” system assigns colors to specific grammatical forms. The psuedo-3D appearance of the buttons looks a little dated to my eye but the perceived affordance may be necessary for some users. The program itself is highly customizable.

To create a message using the different boards, a user would navigate through the system and click on each component in turn until they were finished. For the purposes of measuring message complexity, each of these steps counted as one “navigational unit.”

A simple request for a sandwich might look like this in a context-driven environment (for an utterance of four navigational units):

The same request in a word-based environment would look like this (for an utterance of three navigational units):

My wife’s student’s communication level is context-dependent. However, the navigation boards available for context-driven communication were too complex for him to use and many of the topics simply weren’t relevant. (He would never use either of the above examples because he doesn’t eat solid food — all nutrients are provided through a gastro-intestinal tube.) To get around some of these issues, she programmed a customized board based on his particular abilities and interests.

Some of these modifications were fairly extensive. Since her student had no understanding of grammatical structure at this time, she simplified the color scheme so it only used three colors: orange for the “back” button, blue for any folder that could be opened, and gray for any item at the bottom of a decision tree. She also tightened up the button groupings to reduce difficult eye movements and eliminated any buttons that would appear “underneath” the back button to reduce navigation errors. Finally, she set the dwell time between 8.5 and 9 tenths of a second — the effective “window” for reading the student’s gaze accurately.

.

Customized Communication Board

.

The student was able to use the system for about 2 1/2 weeks in late Spring 2011 and for one week in Fall 2011. During the trial period, the student was able to use the twelve-button screen for several language functions, including basic greetings, requests, yes/no responses, exclamations, expressions of physical state, and even a few jokes (knock, knock jokes that my wife programmed into the computer). The range of communication partners included school faculty and several family members.

For casual observers, the student’s performance using the device was revelatory. One teacher who overheard the student working on a craft-related activity stated simply, “Wow, there’s a person in there.”

Although, it might seem obvious that such a tool would be beneficial for this particular student, the services were not deemed “medically necessary” and the initial request for Medicaid was denied. The evaluator felt that there just wasn’t enough evidence showing independent use of the system to create novel utterances. (Attempts to include some peer-appropriate language may have backfired when the evaluator dinged the student for overly frequent use of the phrase “smell ya later.”)

Another, longer trial was suggested.

2nd Trial

The next loaner arrived in April 2012 and my wife was determined to gather more quantitative data and provide as much documentation of the second trial as she could. Each of the student’s statements during the trial period were marked down and evaluated for complexity (number of navigational units or levels), conversational turns (the alternations or volleys between two speakers), and functions. Functions include descriptions (items, past events), requests (actions, information, objects), responses to requests, social devices (spontaneous calls, exclamations, greetings) and statements (emotions, future events, personal information, opinions). After four-weeks, there were 265 individual utterances available for analysis.

A few initial findings:

• The student’s accuracy of responses to yes/no questions increased to 80% using the eye tracking device in conjunction with the SGD (compared to 20-30% on the eye gaze board).
• The student’s ability to look at an item on command improved to 85%.
• The student was able to comprehend all of the noun and verb phrases programmed into the device.
• The student demonstrated comprehension of the following:  categories, colors, shapes, sizes, actions words, possessives, time words, words denoting quantity, pronouns and wh-questions.
• The student spontaneously accessed the machine to call attention and participate in conversations with a variety of adults and peers.
• The student combined multiple symbols to create a message and often used one symbol in novel ways. For example, he would use “bye” to indicate that he wanted to stop an activity.
• The student demonstrated the ability to repair conversational breakdowns. After an unintended response, he would often use the method of multiple “clicks” on a word to emphasize his correctly intended response.

During the trial period, the student gradually shifted from single-level utterances to more complex navigational structures. By the second half of the trial, 61% of his utterances used a combination of symbols and the average length of utterance increased from about 1.6 navigational units during the first two weeks of the trial to over 1.8 navigational units in the second two weeks. A basic MS Excel t-test performed on this metric suggests that this change was significant.

.

Distribution of Utterances by Navigational Units (1 vs > 1)

.

Distribution of Utterances by Navigational Units

.

The mean score for Half 1 (M=1.605 SD= 0.727, N= 119) was significantly smaller than the mean score for Half 2 (M=1.836, SD=0.822, N= 146) using the two-sample t-test for unequal variances, t(261) = -2.42, p <= 0.016. This implies that the student has the attention, memory, and problem-solving skills to use a SGD to achieve his functional communication goals.

t-Test: Two-Sample Assuming Unequal Variances

	Half 1	Half 2
Mean	1.605	1.836
Variance	0.529	0.676
Observations	119	146
Hypothesized Mean Difference	0
df	261
t Stat	-2.42
P(T<=t) one-tail	0.008
t Critical one-tail	1.651
P(T<=t) two-tail	0.016
t Critical two-tail	1.969

Interestingly, many of the student’s more complex utterances were in conversations with peers — pre-teens with no training in speech and language communication. The student also increased the number of conversational turns per topic over time and, as with conversational complexity, his performance was better with his peers. He had longer conversational “volleys” and used many longer strings of symbols than his conversations with adults.

.

Navigational Units Comparison by Listener

Listener	1	2	3	4
Peer	45.6%	40.5%	8.9%	5.1%
Professional	46.2%	36.3%	16.5%	1.1%

.

Conversational Turns Comparison Over Time

Half	1	2	3	4	5	6
1	59.5%	25.0%	9.5%	4.3%	1.7%	0.0%
2	52.8%	27.1%	10.4%	5.6%	3.5%	0.7%

.

Conversational Turns Comparison by Listener

Listener	1	2	3	4	5	6
Peer	53.9%	23.7%	11.8%	6.6%	3.9%	0.0%
Professional	55.6%	27.8%	9.4%	4.4%	2.2%	0.6%

.

While there is no doubt that this technology would prove incredibly beneficial in this situation, the strict rules surrounding Medicaid requests makes the outcome difficult to predict. By carefully documenting the results of this second trial (and including some awesome tables and charts), my wife hopes to tip the scales in her student’s favor. The report was mailed yesterday so cross your fingers. As my wife’s student might say (through his technology-assisted communication device): “Let’s get this party started!”

UPDATE: The request was approved on June 22, 2012. There is some hard work ahead but this is a big hurdle to clear. Congratulations and good luck to everyone involved!

The Wall Street Journal recently published a table of income and unemployment data  that presented pay and employment rates for various college majors. The original study by Georgetown University’s Center on Education and the Workforce contained enough additional details that I thought it might be worth trying to incorporate the information into a Tableau visualization.

After a little data massaging, I created charts for both the high-level fields of study and the more detailed individual majors. Each level contains unemployment rates, income levels, and popularity of major measured by number of enrollees.

One of the first things you notice is that, despite frequent claims to the contrary, college graduates with a degree in Education have the lowest median earnings overall. The Education field also has the narrowest range of income and includes four of the ten majors with the lowest median earnings. On the plus side, fifteen of the sixteen Education majors have (or had at the time of the study) unemployment rates below 5.5% — the weighted average rate of unemployment for all majors in the study.

Graduates with an Engineering degree have the highest median earnings overall and a relatively low unemployment rate compared to other disciplines. In addition, seven of the ten majors with the highest median earnings were found in Engineering.

Other majors with good earnings potential included the usual suspects (Computers & Mathematics, Health, and Business) while the best employment prospects were found in Education, Health, Physical Sciences, and Agriculture & Natural Resources.

As for individual majors, the winners in my completely fictitious categories are as follows:

• Most Popular -  Business Management & Administration takes this category with nearly 2.8 million grads holding this degree. The next two majors in line (also in the Business field) weren’t even close — trailing by over a million people.
• Best Prospects -  Actuarial Science beat out four other fully-employed competitors by coming in with a median income of over $80K.
• Worst Prospects -  Clinical Psychology tops this category with an estimated unemployment rate of nearly 20%. Yikes! I also noticed that a number of other majors in the Psychology field had unemployment rates above 10%, which means that intra-discipline career changes for people with this major would be difficult.
• Most Deceptive - The “winner” here is Architecture, an outlier with the lowest median earnings and the highest unemployment rate of all of the Engineering majors. For this category, I wanted a relatively popular major with an uncommonly high unemployment rate … the kind of major that churns out grads and then strands them in the unemployment line. An educational Judas, if you will. (Full disclosure: I have an Architecture degree, but I can’t say I wasn’t warned.)
• Hidden Gem – I’m going to call this one a tie between Petroleum Engineering and Pharmacy Pharmaceutical Sciences & Administration. Petroleum Engineering has a slight edge on median earnings ($127K vs. $105K) but the Pharma major has a lower overall unemployment rate (3.2% vs. 4.4%). You probably can’t go wrong with either one but keep on eye on the horizon … Petroleum Engineering is notoriously dependent on the boom/bust cycles of the oil and gas industry while workers in the pharmaceutical industry are facing major changes as companies try to adjust to globalization and increasing costs of product development. 

The sheer ubiquity of the Microsoft Office suite has created a cottage industry around the evaluation and critique of its bundled applications. Microsoft Excel, with its attendant realm of spreadmarts and shadow databases, seems to draw the most negative attention from the business world (particularly IT) but PowerPoint isn’t far behind.

At various times, the world’s leading presentation software has been banned by CEOs of some of the world’s largest companies, called an internal threat by the U.S. military, and served as the driving force behind the establishment of the Anti-Power Point Party — a Swiss political party whose only stated goal is to rid the world of boring presentations. It has even been suggested that the “chronic use of PowerPoint” at NASA helped obscure critical information that might have prevented the 2003 Columbia space shuttle disaster.

How has a little presentation program like PowerPoint earned the ire of so many people? Sure, the tool has its flaws (discussed here, here, and here) but are these shortcomings really the cause of all the world’s slide show ills? Well, yes and no.

Generalized applications like those found in the Microsoft Office suite help companies hold down costs while providing their workforce with a fairly decent toolset. However, once these programs are in place, there is rarely any business incentive to provide additional training or purchase more specialized applications for more complex tasks. This leaves users in a bind. Either they can sit around waiting for more instruction and more powerful tools or they can start experimenting with the tools they already have available.

Like its close cousin, Excel, PowerPoint suffers from the fact that most people end up using it for tasks that it was never designed to do. In the case of Excel, a simple accounting application has become the de facto database and analytics package for most businesses while, with PowerPoint, a basic slide management tool has supplanted lectures and written reports to become their sole information delivery platform.

You would think that PowerPoint would be well-suited to this role. People seem to prefer multimedia presentations over standard lectures and studies in dual-coding theory suggest that they retain more information from presentations that have both verbal and visual content. However, because most speakers don’t really emphasize the full visual capabilities of PowerPoint, their presentations become a combination of verbal and textual content … and retention of information presented in this format may be much worse.

The problem is that cognitive process of creating a presentation in PowerPoint is a lot different from the cognitive process of watching a presentation in PowerPoint. Speakers get so involved in the preparation of their slide deck that they rarely give much thought to how it will be received by the audience.

Max Atkinson sums it up:

“PowerPoint makes it so easy to put detailed written and numerical information on slides that it leads presenters into the mistaken belief that all the detail will be successfully transmitted through the air into the brains of the audience. “

This assumption fails because:

“… the audience’s attention is split between (1) trying to read what’s on the screen at the same time as (2) listening to and following what the speaker is saying and (3) looking repetitively from speaker to screen and back again.”

Simply adding a graphic element to a text slide doesn’t necessarily improve knowledge retention, either. Researchers have found that the use of unrelated pictures in a presentation (think clip art) can actually distract the audience from the main content and interfere with overall learning. This is because people end up paying too much attention to the non-essential material on the screen and not enough to the text or narration.

A more effective technique involves the use of custom-designed images that relate directly to the concepts being presented. Even very complex topics can be tackled using such a combination of pictures and text and the level of information recall is much higher. However, this approach requires a better understanding of how people absorb information and not everyone will have the time or inclination to learn the basic principles of multimedia design.

The realities of human learning would seem to suggest that a presenter error on the side of simplicity, but that approach comes with its own set of pitfalls. For example, some experts say that you only use one slide for every 2-3 minutes of speaking while others suggest that you should never use more than 2-3 sentences per slide. Doing the math, this means that the “ideal” PowerPoint presentation would deliver no more than 1 to 1.5 sentences to the audience every minute and an hour-long presentation would have a maximum of 90 sentences. Even an adult with below average reading skills can tackle that amount of text in about 10 minutes. Expecting people sit through such a glorified guided reading course is a recipe for boredom.

Some executives respond to this kind of presentation bloat by imposing an upper limit on the total number of slides … say six or so. While this directive cuts down on the overall size of the presentation it also starts to have a negative impact on the content. To meet this restriction, presenters are either going to try and cram more information on the few slides they have available (making their presentation incomprehensible) or dumb down their presentation entirely (making it irrelevant or even ridiculous).

Therein lies the dilemma. Some say that creating a meaningful presentation in PowerPoint is impossible:

“There is simply no way to express precise, detailed and well-articulated ideas or subjects through Powerpoint.”

Others say that the tool is perfectly fine and that any fault lies with the user:

“Is PowerPoint bad? No, in fact, it is quite a useful tool. Boring talks are bad. Poorly structured talks are bad. Don’t blame the problem on the tool.”

My own thoughts tilt towards the idea that the tool has been badly misused and its reputation can be redeemed through proper use. Many people have written extensively on what you should and shouldn’t do with your PowerPoint presentations (here’s one) but I have distilled my own thoughts on the subject down to three basic rules or guiding principles (with exceptions, of course):

1. Don’t use any text – That’s right, you heard me people … none. PowerPoint is a visual medium and should only be used for visual images. You’re supposed to be telling a story, not writing a grocery list. Simply putting your speaker notes on screen is a cop out and will leave your audience squirming in their seats after the first five bullet points.  Yes, you can create an outline to help organize your thoughts, but by the time you’re done developing your presentation, these blocks of text should be gone. Exceptions: Every visual medium uses some text on occasion. Things like titles, section breaks, tables, end notes, and explanatory text on charts are all welcome in moderation … but, if you strive for a slide deck that is 100% text-free, you might actually achieve something that is 80% text-free, which is way better than 90% of PowerPoint presentations out there.



2. Only use images or videos that you create yourself – As you struggle with the content of your presentation, it is always tempting to add a little cartoon, GIF animation, or random stock photo to spice things up. Don’t do this. Your presentation should be tailored to deliver a specific idea to a specific audience. Adding someone else’s work to your presentation – even a picture from your company’s own brand library – is just a distraction. Build your own charts, draw your own diagrams, and create your own videos. You will be rewarded with a presentation that is consistent and perfectly suited for your message.Exceptions: If you are you are truly creatively challenged, find someone else who can help you visualize your ideas. Don’t appoint a committee to the task, however, since you want to maintain a consistent visual language.



3. Focus on your delivery, not your handouts – Using PowerPoint to display pictures and graphs that support your presentation is good … using PowerPoint as a crutch to help you get through your talk is bad. Memorize what you want to say and prepare notes that you can use for reference while you are speaking. The audience should be getting a well- delivered presentation from someone who is organized and confident, not the half-formed thoughts of someone reading from their slide handouts. Exceptions: Seth Godin recommends creating a written document that complements your PowerPoint presentation and handing it out after you’re done speaking. This document shouldn’t substitute for adequate preparation but it should support your key points and provide additional details that help your audience understand the topic.

P.S. For a PowerPoint presentation of this post, click here.

I was looking at an online timeline development application from xTimelines but I couldn’t get the embedding feature to work on this site. A quick search for similar tools led me to my alma mater, of all places, where they had done a nice evaluation of alternative timeline applications.

Here’s a sample from Timerime:

I also liked the Timeglider application but I thought the navigation was a little less intuitive. One interesting facet of this tool is that the company has teemed up with the New York Times to create an online application that generates a timeline of headlines for any search string. This is a powerful concept that takes advantage of the NYT data API … something I’ll have to explore further.