GARY C. RAMSEYER'S FIRST INTERNET GALLERY OF STATISTICS JOKES BY TOPIC

"79.48% of all statistics are made up on the spot." - John A. Paulos

*Thanks to Bill Weaver for this quickie!

Did you know that there are _three_ kinds of statisticians--those that can count and those that can't.

*A big thank you to a fellow Hawkeye, John Creyer, for a great chuckle.

A friend was bragging to a statistician over coffee one afternoon how two-day volatility in the stock market had treated his holdings rather kindly. He chortled, "Yeah... yesterday I gained 60% but today I lost 40% for a net gain of 20%."

The statistician sat in horrified silence. He finally mustered the courage and said, "My good friend I'm sorry to inform you but you had a net loss of 4%!!!"

* My little tale above illustrates how pervasive innumeracy is in our society. Always remember, "Percent of What?"

A prisoner had just been sentenced for a heinous crime and was returned to his cell. An inquisitive guard could not wait to ask him about the outcome.

Guard:"What did you get for a sentence?"
Prisoner: "I could choose life or 100 years."
Guard: "And what did you choose?"
Prisoner: "Well, life, obviously. Statistically speaking that is shorter."

*This convict obviously knew a little about statistics but was lacking in common sense. Thanks go out to Coen Bernaards from UCLA for sending this one my way.

A reader of this Gallery sent me a very amusing story.

He took advantage of one of those online offers...a free credit report. He was delighted to learn that his credit rating was better than 100% of those who had received such a report...which obviously included himself. "Whoopee!!!, he exclaimed, my credit rating was better than MY OWN!!! It just doesn't get any better than that."

*Harley(I need your last name), thanks for this cute anecdote. This does hit on a point that I have been frequently asked about. Can a person's score in any distribution fall at the 100th percentile or more precisely can his score have a percentile rank of 100? If you subscribe to classical test theory, the answer is technically NO. Suppose that person A had a top score of 23 on a 25 item test with the remaining 49 other students scoring below 23. Then assuming the scores are continuous, person A's true score would be between the real limits of 22.5 and 23.5. The only way that person A's score would have a percentile rank of 100 would be if his true score was between 22.5 and 23. Since it is just as likely that his true score is between 23 and 23.5, we generally compromise and assume 1/2 of his score is between 22.5 and 23 and the other half is between 23 and 23.5 (a whacky assumption but more plausible if you had several scores of 23). Thus the percentile rank of person A would be (49 + (1/2 x 1))/50 = 49.5/50 = 99. Most modern authers subscribe to the above line of thinking but as we all know, statistics is heavily laden with many assumptions.

The secretary of defense gave the president his daily briefing. He concluded by saying:
"Yesterday, 3 Brazilian soldiers were killed."

"Oh No!" the president exclaimed. "That's Terrible!"

His staff was stunned at this display of emotion, nervously watching as the president sat, his head in his hands.

Finally, the president looked up and asked, "Just how many is a brazillion?"

*Here is another example of the ubiquitous innumeracy that is gripping this country. Thanks to my good friend Merle (Pearl Diver) Howard, an emeritus Professor of Speech Pathology at Illinois State University, for forwarding this little story to me. I wonder if the White House could use a good speech pathologist as a consultant these days?

50% of marriages end in divorce. Thus if you don't file for divorce, your wife will.

*This is a cute little variation of all the 50-50 jokes. But wait a minute! This says the probability of any marriage ending in divorce is one. Sorry I don't have an attribution on this one.

It is a well established fact that Statin Drugs ($27.8 billion sales worldwide in 2006) lower the BAD cholesterol level by 20-50% in most individuals upon continuous treatment. In fact, in a recent TV ad for a certain Statin Drug endorsed by Dr. Robert Jarvik (the artificial heart man), it was stated that this drug "reduces the risk of heart attack by 36% ... in patients with multiple risk factors for heart disease but no heart disease."

"Zowie, exclaimed a friend watching the TV ad , this is comparable to penicillin. Since they called these drugs Statins because they are 'Statistics Insured', why not add this to our drinking water supply and let everybody from babies on up benefit??"

*Now before you petition your city council or go out and buy stock in this Drug Company, lets take a very careful look at this 36% figure which the Company itself tagged with an asterisk in the original report. Here is another prime example of innumeracy and number enhancement at its best. In smaller print the Company explained that this means in a large clinical study that lasted 3 and 1/3 years, for every 100 subjects treated with the drug, 2 people had heart attacks, as compared with 3 heart attacks in a group of 100 subjects that rececived a placebo. In other words, for every 100 subjects who took the drug, 1 less subject was spared a heart attack(i.e. 1/3=33% ≈36%). But this is a spurious percentage so we need to put this in terms of a relatively unknown but useful statistic called the NNT. I want all my readers to etch this in their mind. The NNT is the number needed to treat for one person over the expected number to benefit!. In our case the NNT is 100 ... that should have a very sobering effect on the exhuberant 36% reported earlier! Now to show you some other examples of this NNT: (a) NNT = 2 to cure a stomach ulcer with antibiotic cocktail in one year of treatment (this is a great NNT); (b) NNT = 16-23 to prevent one heart attack with a Statin in people that have had a heart attack or other signs of heart disease (most likely group to beneft from a Statin); (c) NNT = 500+ to prevent death or serious medical condition with a Statin among patients without heart disease that have a single risk factor like high blood pressure (Probably should not be on Statin at all); (d) NNT = 1000+ for Avandia, a drug for controlling blood sugar to prevent heart disease or heart attack (Even companies admit that this NNT is way too high for any benefits). Of course, along with the NNT, one must also consider the side effects of the drug in making any decision about long term usage. In the case of the Statin we mentioned before, it is very interesting that the Company essentially uses the NNT to downplay these side effects like muscle aches and memory problems. It states that only 1 in a 100 will experience side effects of some sort. WONDER WHY THE COMPANY CAN'T EMPHASIZE THE NNT FOR BOTH ISSUES? Much of this discussion is based on an article in the January 28, 2008 issue of BusinessWeek magazine. I urge all students of statistics and actually everyone to read this article online Jan 28, 2008 BusinessWeek This is one of the most illuminating articles on Drug Research that I have read and covers other key issues of drug effectiveness. By the way, the name "Statins" did not come from "Statistics Insured"... this was just part of the humor.