Stats Made Easy

Having had enough of statistical tea-testing (inside joke), I bought two brands of beer this week — Summit Grand “pilsener” (Bohemian style) and Schell “pilsener” — for an impromptu test of my wife’s tasting abilities. She comes from 100 percent German stock consider beer as dear as mother’s milk. Thus, I was not surprised she got every one of these four combinations correct (“her answer in quotes”), even though I revealed nothing about the sources, saying there may be from one to four beers:
– Grand vs Grand “same”
– Grand vs Schell “different, but I don’t prefer one or the other”
– Schell vs Grand “different, but I don’t prefer one or the other”
– Schell vs Schell “same”.
The Grand beer is clearer, less expensive and offers a twist-off cap, so my wife and I will stay with this brand made in my home town of Saint Paul, Minnesota (Schell is made in New Ulm — a German settlement in the southwest of the State).

I would appreciate other ideas for a simple, but scientifically valid (?) taste-off like this for beverages or foods. My very cursory internet search brought up an article by James Fallows in the e-zine “Slate” titled Booze You Can Use, Getting the best beer for your money, but I do not necessarily advocate its methodology. I do happen to like Sam Adams a lot, which the Microsoft employees doing the tasting rated as best. Well before 1984, when Jim Koch founded The Boston Beer Company and the idea of a microbrewery, a chemist friend of mine snobbishly proclaimed that he only drank the “champagne of bottled beer,” Miller, and not the much cheaper (at that time) Old Milwaukee brand. I’d just taken a marketing class for my MBA that revealed that beers of this era (late 1970’s) were all essentially the same, but advertisers duped drinkers into paying more for “premium” brews. The chemist refused to believe this, so two of us chemical engineers set up a beer-tasting contest for a Super Bowl party. First we all sipped 10 brands of beer straight out of the bottle (perhaps a bit too much!). The chemist rated Miller top and “Old Swill-waukee” bottom. Then we repeated the test with beers poured blind to the tasters. The chemist unknowingly declared Old Milwaukee his favorite and, you guessed it, Miller the worst tasting. Of course by then he was good and drunk, but the point was made. Anyways, we sure had a lot of fun pretending to have a taste for beer!

Is there method in some of the March Madness of college basketball playoffs? Point shaving occurs in five percent of games with a spread higher than 12 points according to self-styled forensic economist Justin Wolfers in an interview with Jeff Mason. Wolfers is a University of Pennsylvania (Wharton School) professor who teaches the world’s only known course on the Economics of Sports Betting Markets. He extracted 9,244 games where teams were favored by 12 points or more from 44,120 college basketball games played between 1989 and 2005. Significantly, 46.2 percent of strong favorites won outright but failed to cover, which Wolfers thinks is a clear indication of point shaving. All a fellow has to do is miss on basket when his team leads by 10 points to come up short on the point spread of 12, thus making good on a big money bet. All this can be very tempting, so much so that the NCAA Gets Proactive On Gambling by inviting the FBI to talk with teams such as those going into the 2006 finals. Evidently players must get a lot smarter if they want to beat the system. Perhaps they had best pass on the basket-weaving classes and take a course on statistics!

According to statistics from a recent survey by America Online and Opinion Research Corporation cited in the New York Times*:
– 41 percent of the respondents check their e-mail in the morning before going to work
– 25% said they never go more than a few days without checking e-mail
– 60% check it on vacation
– 5% looked at e-mail in the bathroom.
That last one is where I draw the line!

Based on the title of her book, time-management expert Julie Morgenstern evidently advises that to avoid too much stress due to doing work at home one should Never Check E-Mail in the Morning. However, according to Steve Pavlina once you check email in the morning (before work?), you should not check it again until the end of the day. That seems extreme to me, but I like Steve’s idea to experiment on how infrequently you can check it without causing problems, that is, assuming feel that your email has gone out of control. I wonder about this myself. The thing is that I love to read and I like to write, so for me, email really may be addictive. I will know this is so the first time I find myself doing it in the bathroom!

*(11/20/05 article by Lizette Alvarez titled “Got 2 Extra Hours for Your E-Mail?”)

The Twin Cities got dumped on today with nearly a foot of heavy, wet snow. The weight of it snapped trees branches around my yard and left our Stat-Ease office powerless due to downed lines. I cannot say that we Minnesotans were not warned, but just before the storm hit, I heard one television weather man, a new guy giving his first forecast, say that the snow would fizzle due to the low pressure sucking in dry air. This reminds me of the movie “Weather Man” a dark comedy starring Nicolas Cage, whose character at one point before going on television decides to get the forecast from the real meteorologist behind the scenes:
> Meteorologist (M): “You should say, ‘We might see some snow…’ “…but it might shift south, miss us.'”
> Cage as Chicago TV weather guy: “I can say it, but I sort of wanna understand it. Why is it?”
> M: “Well, it’s Canadian trade winds.”
> Cage: “Behind all of it?”
> M: “Yeah, this will get pushed by wind out of Canada.”
> Cage: “So what’s it gonna do?”
> M: “I don’t know. It’s a guess. It’s wind, man. Blows all over the place.”
It doesn’t matter how variable snowfall can be for any given area, the new weathercaster in the Twin Cities television market will be in the doghouse now for a long time because he got this biggest storm of the winter wrong. As the move “Weather Man” says in its tag line: “In life, accuracy counts.” On the other hand, in our business of industrial statistics for predictive modeling we allow ourselves an out by this disclaimer: “Statistics means never having to say your certain.” This is lot more realistic!

PS. FYI, here’s a link to a forecasting model devised from neural network analysis of historical records that allows you to predict the ratio of snowfall depth to liquid water, a vary tricky property: University of Wisconsin – Milwaukee (UWM) realtime snow ratio forecast page .

I just finished off a bottle of Rosso Classic by Francis Coppola, the famous film director of the Godfather series and Apocalypse Now (the movie based on Conrad’s Heart of Darkness). This wine, which I really liked, is a blend of 47% Zinfandel, 32% Cabernet Sauvignon and 21% Syrah from California.

A couple of summers ago I promoted use of statistical design of experiments (DOE) to the Society of Industrial Microbiology at their annual conference. One of the attendees was a professor in the Department of Viticulture & Enology, University of California. What a great job!

I first became interested in the potential of DOE for wine-making when I came across some intriguing case studies by Douglas Montgomery in his textbook Design and Analysis of Experiments. A recent writeup on Professor Montgomery at Arizona State University (ASU) Regents’ Professor: Douglas Montgomery mentions that he’s “delved into aspects of chemical engineering in a project to develop an efficient wine-making process for an Oregon winery.” I believe his specialties are Pinot Noir and Pinot Grigio. The sensory evaluation of wine involves smell and taste.

I find these intriguing facts about human sensations at International Journal of Food Science & Technology

— There are around 10,000 taste buds, each of which contains 50–100 cells detection sweet, sour, salty, bitter and umami (produced by monosodium glutamates and the like).
— Humans can distinguish around 10,000 chemicals with upwards of 100 million olfactory receptors!

If you really want to enjoy an flavor sensation (combination of taste and smell), add cheese to the wine and make a pairing. According to Discovery News Brief “the first attempt to study wine and cheese pairing using a controlled environment and a scientific experimental design” found that the ideal combinations are Gewurtztraminer with a creamy blue cheese, unoaked Chardonnay with a semi-soft washed rind cheese and a blush wine with an aged cheddar. My suggestion is that these findings require confirmatory experiments!

I just returned from a week of surf and sun — a welcome respite from the coldest spell this winter in Minnesota. I went from -15 to 85 F from takeoff to touchdown over the four hour flight south — a span of 100 degrees! Mazatlan is home to the family of a young lady named Clarissa who spent a year with us as an exchange student — joining my three daughters in our residence in Stillwater. (You cannot imagine the pileup in the bathroom each morning before school!) Clarissa frequented the local health food store where, if I had not forbidden it, she would have purchased a scary collection of unregulated medicines for losing weight. It’s bad enough that the US Food and Drug Administration (FDA) provides no protection against the snake oils and such that are peddled to naive consumers, but south of the border in Mexico it is far worse. There, for example, many desperately ill Americans find “alternative” medicines and treatment, although as a colleague at Stat-Ease pointed out, either it is medicine or the alternative. I see in Mexico an unabashed belief in the supernatural, which I suppose is just as prevalent here in the USA, but relegated to tabloid and cable television features about haunted houses. Coincidentally I brought along for beach-reading the latest issue of American Scientist which included an article titled “The Cognitive Psychology of Belief in the Supernatural” that explains why humans harbor an innate faith in the unreal. I found this quote enlightening:

“…our brains have evolved so that science eludes us but religion comes naturally.”

PS. This issue of American Scientist also features an article telling why “Three statistical strategies —- replicating, blocking and modeling —- can help scientists improve accuracy and accelerate progress.” I agree! See Volume 94, Number 2, March-April 2006 .

James N. Cawse of GE Global Research emailed me this week with this question.

“Mark, has anyone developed a demonstation/experiment suitable for high school students that illustrates a Latin Square design? I’ve been asked to give a high school chemistry class on combinatorial chemistry; as I thought about it, the simplest “combinatorial” type design is a Latin Square. It actually has a chance of being understood because of the current craze for Sudoku.”

My response was:

“James, Funny you should mention Sudoku and Latin Squares because this morning I was thinking how the same structures apply to Latin Hypercube Designs (LHD) that are popular for DOE on computer sims, for example ones based on finite element analysis that GE uses for designing jet and power turbines. (An aside — I just got the Jan-Feb issue of American Scientist featuring in their Computer Science column an article titled “Unwed Numbers” on the mathematics of Sudoku. This name stems from the Japanese firm Nikoli who called these puzzles “the numbers must be single” in the sense of being unmarried.) To answer your question, no I don’t know of a classroom experiment that illustrates the Latin Square structure.”

If any of you StatsMadeEasy blog readers knows of good in-class chemistry experiments that illustrate Latin Square or other principles of DOE, post a comment.

PS. I see that Wikipedia offers a very extensive entry for Sudoku that includes this comment about the mathematics of it:
“A valid Sudoku solution grid is also a Latin square. There are significantly fewer valid Sudoku solution grids than Latin squares because Sudoku imposes the additional regional constraint. “

As a dues-paying member of American Institute of Chemical Engineers (AIChE) I got my Chemical Engineering Progress (CEP) magazine today — the Feb 2006 issue. I see in the article “Designing Experiments for the Modern Micro Industries” that author Phillip H. Williams claims that in his semiconductor industry the processes are so complex that engineers canNOT assume that the sparsity of effects principle* rules. He then supports this contention by showing numerous three-factor interactions (3FIs) from Minitab software analysis of a full 32-run two-level factorial on 5 factors. After realizing that Table 1 showing response data got out of standard order (1,2,3,…19,20,21,24,22,23,25,26,27…32), I got Design-Expert version 7 software to agree with Williams’ results. However, from the Box-Cox plot it is evident that an inverse square root transformation helps. Oh, and by the way, Williams uses the relatively risky p-value of 0.1 as the cut-off for significance. As a practical matter I would say that main effect predominate as predicted by sparsity of effects. However, it appears Williams does have some basis for saying that this principle fall down in his case, which produces a number of apparently significant 3FI’s. Nevertheless, I am not swayed (as he is) from the advice (quoted from the article) that “if you have five factors … never do the full factorial since the 2^5-1 is a resolution-five design.” This still makes sense to me — why do 32 runs when 16 will normally do. I will email my DX7 file to anyone interested in playing with this data.

*From Wikipedia: “The sparsity of effects principle states that a system is usually dominated by main effects and low-order interactions. Thus it is most likely that main (single factor) effects and two-factor interactions are the most significant responses. In other words, higher order interactions such as three-factor interactions are very rare.”

Being at Excel Center rink to see the Minnesota Wild clobber the Kings of LA at their game last week in Saint Paul and then watching the USA women hockey team beat up their first two foes, Switzerland and Germany, at the Turin (aka Torino) Olympics got me thinking about the technology of hockey sticks. I have fiddled with small scale hockey sticks for in-class experiments* but I wondered if anyone had posted technical detail on the real thing, so I searched the internet. I was very surprised to get numerous hits on global warming! According to BBC News “The hockey stick was a term coined for a chart of temperature variation over the last 1,000 years, which suggested a recent sharp rise in temperature caused by human activities.” But then I came across numerous web entries disputing the “hockey stick” as an artifact of principal component analysis, or PCA, which evidently due to improper normalization procedure tends to emphasize any data that do have the hockey stick shape, and to suppress all data that do not. For one detailed opinion on all this, see www.john-daly.com/hockey/hockey.htm. Being a native of what we Minnesota residents often call the “State of hockey” and a fan of statistics, I find this all very interesting. I do not care to get snowed under with arguments about global warming, but I have put off plans to buy new hockey skates this year — it was too warm this year to produce good ice at the local outdoor public rink. However, the record temperatures here in Minnesota did not stop the 27th Annual International Eelpout Festival this weekend at Leech Lake. The sheriff did prohibit fishermen from driving their SUVs out on the ice. Earlier this season a bunch of ice anglers lost their vehicles after parking them too close together. SCUBA divers fished the SUVs out from under the ice. Perhaps if the windows were left open they might have trapped some fish hoping for a joy ride.

*See “Tabletop Hockey Meets Goals for Teaching Experimental Design”

The article “Corporate Blogs for Technology Businesses”* published by www.productmarketing.com says that according to Technorati, a popular ‘blogosphere’ search engine, a new blog is created every second. Here are other stats from the article:
– 16% of U.S. adults (32 million) are blog readers, a 58% increase over 2004 (from “The State of Blogging” by Pew Internet & American Live Project)
– 6% of American adults has created a blog (11 million people)
These figures stem from early to middle 2005, so by now they may be considerably higher based on what I am seeing in the media — a huge buzz on blogs.

*Authored by David Meerman Scott, whose blog is www.webinknow.com