Aces Full of Links

James showed me the Colbert clip and my favorite part isn't even the end -- I think the best part is when the Congressman says he can't think of a more appropriate building to house the ten commandments. A little scary insight there.

You're sneaky; I didn't even see you make that comment.

Maggie is right - that's the funniest moment. The irony fairly drips off the video. Clearly, there are more important concerns than actually having the 10 commandments in an appropriate venue... like pandering to religious voters.

I was in tears when I saw that on the show the other night. "You can't think of ANYplace more appropriate?"

I mean, Colbert can make most people look foolish when he interviews them, but he didn't have to resort to any of his usual tricks to accomplish that here.

The graph of your weight is using a common deceptive tactic. The y-coordinate (your weight) starts at 166 (at the moment). It should start at zero. What you're doing is also known as a "zoom", which has the effect of making the change look much bigger than it is.

I know, I'm being a jerk. You're using the graph to motivate yourself. But, I can't resist mentioning it, since I've been trained to be skeptical of graphs.

You're not being a jerk. Unless, er, you're trying to be. In which case I'll let you be the judge of that.

Deception implies an agenda, and I have none. The graph is generated for me by a website CalorieCount.com. The graph extents are generated automatically by the site, and I have no control over them. I'm assuming that they generate the extents based on centering the data for maximum viewability and detail, not to deceive anyone.

I don't mind the comment (really I don't), but to give some context just so you don't think I'm a bumpkin who doesn't think about data representation, I have worked for 12 years on NSF-funded research projects creating software dealing with the use of dynamic graphing of functions in math education. Does that makes me any smarter than anyone else? No, but it does mean I've had to think about the implications of graphs once or twice over the course of writing and designing a few graphing applications. (Of course, now you're thinking, "Aha! He's a graph expert trying to fool us with his subtle knowledge of graphs!!!! ;) )

Having the origin off the bottom of the graph accentuates the apparent slope of my weight loss. It would be more shallow if the axes were rescaled to include zero. Or, it would be huge and have the exact same apparent slope, using the same aspect ratio and a lot of blank space at the bottom. However, anyone that interested can calculate the actual slope, since the units are given.

So, I would encourage all of you to calculate the slope of my weight loss so I don't pull the wool over your eyes with my deceptive data representation.

BTW - my average weight loss rate is just over 1 lb a week, if anyone cares, which I'm sure someone does. It's been fairly steady since I got back from Texas, even with my indulgence at Opa last week. That's because I reduced my caloric intake before and after the meal. I have had some spikes in my weighings, but the running average smooths them out to reveal the trend, which (to me) is the important thing.

It never fails, even on my day off I have to write somethings about graphs. Pft.

Thanks for caring, I guess.

What's the point of generating a graph that starts at zero? The slope of a healthy weight loss would barely be noticable from that distance and it would resemble a straight line. That's silly.

Our software automatically scales the axes to show only the range of meaningful data. There's no point in displaying beyond the logged values, since there's no information there. (The user can certainly zoom out, though, if they're wondering what a straighter line and more white space would look like.)

Mike and Julie, yes, it is an important point that the data itself gives you the context to determine what is and isn't a meaningful graphical representation.

General rules on what a deceptive graph is aren't very useful because what is deceptive for one purpose may be useful for another.

I happen to think that CalorieCount's approach is useful, because I want to be able to see lots of detail in the Y axis, not lots of blank Y axis (Julie's point). Mike's point is well-made, as there is little chance of me weighing in at zero pounds, so what good is having it on the graph?

If I were gaining weight, however, I could scale the graph from zero (it's the all-important origin) to the upper limit of human weight, which is somewhere around 1400 lbs. (believe it or not). Within that graph I could easily hide my weight gain from you folk, and eat all the ice cream I like. I could gain 20 pounds and you'd hardly notice (it'd be a 4 pixel blip).

Anyhow, for all any of you know, I'm losing weight by having important bones removed, which, as we all know, is cheating and unhealthy.

I think Bill has a sense of humor and didn't intend his post entirely seriously. He's probably just busting my balls, as they say.

And there is no metric for that of which I am aware.

CalorieCount has an agenda to motivate you to lose weight and be happy with tracking your data with them.

Looking at the graph shows a sharp, dramatic slope downward. It implies something much more than is true. In truth, you lost about 13 pounds.

I will grant that the zoom allows you to see the changes in the line better, but it also exaggerates them. In truth, those day-to-day changes are fairly meaningless. Yet, this graph makes them seem significant.

I don't mean at all to pick on your weight-loss here. Losing a pound a week is great progress and in fact most experts don't recommend that you lose much faster than that. Also, I'm way fatter than you and have completely failed to lose weight and keep it off. So, you're a better man than I. But, your graph is a kind of evil that I will not let stand without challenge!

Graphs that start at zero show percentage changes, visually. Thus, they are honest. From a quick check of the numbers, it looks like you've lost about 7% of your original weight. If the numbers were removed from this graph and you were to show it to someone, do you think they would think it implies a 7% drop?

I do have a sense of humor, but not about graphs.

(that was a joke)

If someone's goal is to lose twenty pounds then is a 7% change in total body weight meaningless? I think not. Nor are the day-to-day changes meaningless. A visible downward slope can have a wonderful effect on motivation.

A graph that doesn't start at zero is also honest in this particular context. The information is displayed on a graph that starts at 0 and another graph that starts at 150 is exactly the same and it is up to the reader to draw his own unbiased conclusions from it.

I doubt James is out to impress anyone with a graph he is using to motivate himself, albeit one that is displayed publicly as a way to hold himself accountable.

Sorry Bill. Mike is right. Does a graph have to show percentage changed to be honest? Of course not. Effective weight loss is not based on percentages.

Now I strongly suspect you're having a laugh at my expense, pretending to not understand why having zero appear on a male human bodyweight graph is ludicrous. Simultaneously telling me that the graph doesn't show my percentage weight loss, then calculating it from data you got off the graph. Describing the graph as exaggeratedly dramatic, then telling me my weight loss is great progress.

Whoever you learned this bias for the origin from was playing a joke on you, and you should go slap that person right now and join us over here on the team of people who understand that graphs have a purpose that drives their meaning. People who recognize that there's no such thing as a one-size-fits-all rule for graphing. We will become smarter and smarter and smarter together until we leave these physical bodies of ours and float in space as clouds of energy. With zero weight.

Just in case you're not joking, or if someone else reading this is unfamiliar with why we graph things, graphs are designed to quickly convey information in a way that is hard to see just by looking at the numbers. In the case of this graph, I want to quickly be able to tell changes in the trend of my weight loss.

I need the trend to be as easy to read as possible. If the slope changes, I want to know. That's the purpose of the damned graph. If my weight loss starts to shallow out over a week, I will know that I need to decrease my caloric intake. If my weight loss becomes too steep, that comes with its own disadvantages and so I need to eat more calories.

If I were to include silly numbers on the graph, like zero, or -35, or 1400, it would be difficult for me to spot subtle changes. Thus we see "useless" coincides with the previously used definition of "honest."

Since "dishonest" and "useful" are therefore synonyms, I will strive for the most dishonest graphs possible in the future.

Now I need to go make myself dishonest by feeding the kids breakfast.

"Looking at the graph shows a sharp, dramatic slope downward. It implies something much more than is true."

It would be an "implication" if the y axis were not labeled accurately. Since the y axis is labeled, there's nothing implied. I suppose if you think that the graphed series is the only part of the graph that's important, you could be confused. However, a person who knows how to read a graph (and I'm not talking about "reading a graph skeptically to avoid being misled," I'm talking about "reading a graph to understand the information it depicts") knows that the curve is useless without clearly labeled axes.

"In truth, you lost about 13 pounds." Yes. It is no great acheivement to figure that out - it's why the y axis is there. The *curve* is there to identify trends over time. The curve is showing a downward trend, which is the information James is looking for.

There is more to the trend than just up or down, too. There's also the speed of the weight loss."Downhill faster" or "downhill slower" would also be useful information, and that would be especially hard to see in a spreadsheet as opposed to a graph.

I haven't the faintest idea why % of body weight would be useful. The significance there would have more to do with how fat you were to begin with. I.e. losing 30% of one's body weight wouldn't be nearly enough if you started out at 500 pounds, but it'd be quite dangerous if you started out only a little chubby.

Well, clearly I'm out-numbered and I'm worried that you all see me as some kind of troll now. So, I'll back off and give up. It's really not that big of a deal to me. I think I over-stated or exaggerated my position. I failed to express my point clearly and politely. I don't disagree entirely with what you all are saying. But, I think there is something to my point of view as well (if only I could express it better).

My views on this come from a WPI computer science class I took about 15 years ago, "Human-Computer Interaction". We spent a few days on statistics and graphs and how they are sometimes manipulated for effect. Often the effect is psychological. Also, some of the techniques are in common-use and are essentially accepted practices despite their deceptive qualities.

This context is not the right place to go into all of that. Plus, I'm clearly too rusty on the subject to explain it. Potentially, I am entirely wrong. It wouldn't be the first time.

Please accept my apologies. I will no bring it up again.

Let's laugh about it later over a drink sometime.

To be honest, what we've discussed here, and your point had occured to me when I first started this graph. Not the bit about including zero, but about what would happen over time. As more days are added to this graph, the aspect ratio changes, accentuating the slope. It bothered me that the aspect ration would be changing, not from a deception standpoint, but that it might make it harder for me to judge how fast the graph was changing over time.

It turns out that this hasn't been so much of a problem (yet) because of automatic Y axis rescaling caused by weight loss. It hasn't been completely even, but it's been somewhat proportional. I expect that as I stop losing weight and hit my goal, time will move on and that x axis will get more and more compressed. At that time I may switch back to using Excel.

The true test is what happens when I start gaining weight again. As noted, this graph accentuates the slope. An uptick will be noticeable.

What I likely will do is stop posting the graph when I hit my goal and keep the goal for a sufficient amount of time. Then I will just post my current moving average weight and establish an upper threshold. Breaking the threshold will trigger the diet again, and it's back to the graphs.

At the goal weight, the graph is useful for determining how many calories you need in a day. You pick a number of calories, eat exactly that number every day for two weeks and see what the trend does. From that you can tell how many calories a day you burn. That becomes your maintenance intake.

If you haven't already checked it out, I highly recommend "The Hacker Diet" which is available for free online. It even comes with Excel spreadsheets. I felt it could have been a lot shorter, though.

James,

you've actuaklly had a drink with Bill (just to let everybody know he's not a troll). Think last game day. Both sides of this argument have merits. It is very common to see graphs that are intentionally misleading. Especially in advertising. Not labelling the axis and not including zero would be one way to do it. The other way can have the same effect. I monitor a lot of how our test methods work using control charts. Now in some of these methods a 5% deviation would be a very large change. Let's say the data always returns results that are 495-505 mg/mL. If I included zero on that graph it would be useless. I'd never see a problem unless it were enormous. By the same token if I only graphed 495-505 all I would see is noise. I might miss a true trend and I would certainly miss any outlier. You really need to be careful to apply what you know about what you're graphing and use that knowledge. If you're looking at a graph however it is always best to be skeptical. make sure you look at the axis. I too noticed that your graph had a range that was too small and almost said somethign but decided not to be a dick like Bill (just kidding Bill). If you had only lost 5 pounds it would look the same as if you'd lost 50. What would be better would be something like 20% above you're starting weight was and say 10-20% below your target. then keeping the same axis every week.

Head On is manufactured by Miralus. Their toll free number is 866-647-2587. Call their marketing department to complain.

Head On. It's a stupid commercial.
Head On. It's a stupid commercial.
Head On. It's a stupid commercial.

They show the ad a lot on CN8.