Andrew Hagen

Seemingly tedious issues vex me.

One of the poorest understood and most widely relied upon subjects is statistics. In modern society, the educational system should teach basic statistics to all pupils. Unfortunately, it does not. A primer is online. (*)

Three of the most common offenses are a misuse of percentages, statistical surveys of factual questions, and the employment of web polls for anything other than entertainment.

Every educated person knows something about percentages. The trouble is that so many people do not understand the fundamentals of the concept. Take a person who flips a coin 1,000 times, recording the results of each toss. Let’s say the coin is perfectly balanced. Let’s say that there were 550 heads, and 450 tails. That is, 55% of the tosses came up heads. Now he prepares to toss the coin again. What is the probability of heads in this toss? 55%? No. It is 50%. This is because in probability the past does not matter. There is always a 50% chance of heads, no matter how many heads or tails came before.

Another common, glaring error is the use of a percentage to characterize an upcoming event. Say one sports team is playing another. The first team has ten victories and no losses. The second has ten losses and no victories. A sports commentator might say something like, “The first team has got a 75% chance of winning this game.” Yet, there is no validity in that statement at all. It is unfortunate that this mistake really happens. It happens every day and in all walks of life. Think about it. How many games are these teams playing? Just one. They are not playing four times. Only once. Therefore, at the end of their playing the game, one team will be 1–0 against the other team; while the other team will be 0–1 against the first team. Therefore, the actual chance of victory is 100% in favor of one team or another.

Instead of saying, “this event will probably happen,” say “this event will happen because…” or “this event will happen if….” The misuse of percentages tells your audience nothing of what you really know. Tell your audience why you think what you think.

Sometimes perfectly valid, scientific polls are taken of whether the public believes in certain facts. The trouble is the results seem to have more importance than they really do. For example, imagine a poll that found that 83% of American adults believe Saddam Hussein had weapons of mass destruction when his regime was attacked in March 2003. (I just made this figure up.) Why would it matter? Such a poll has no bearing in any way on whether that is a true statement or a false one. Popular perceptions cannot make a false statement true or true statement false. Perhaps in some cases such a survey could be of help in understanding why other popular opinions are what they are. Surveys of factual beliefs should be carefully framed when presented. Commonly, however, they are presented as if they were in some way intrinsically important.

The last is the worst of all. Web polls are inherently untrustworthy. (†) First, web polls are frequently “freeped.” That is, groups of people who support one position or another, sound or unsound (”We must stop flouridating the water!”), can network online and all take the poll together. This skews the results. Second, polls are frequently sabotaged by those with basic technical skills. They are able to vote more than once. Finally, even if such elementary issues were reckoned with, web polls are open to anyone with a modem, and to no one without. As a result, the poll taker cannot know whether his sample is normally distributed. Most likely, it is not. A scientific poll requires careful design. A web poll requires five minutes and no thought. Draw your own conclusions as to their comparative validity.

Just going through the mantra of “It’s unscientific” is not enough. Major television news networks, newspapers and other sources of information to the public routinely trumpet the results of their dreadful web polls, doing the minuet of it being “unscientific.”

It is time to face the reality. Web polls are not merely unscientific. They are lies. Anyone who believes in a web poll, or uses it for anything other than the object of comedy should be ashamed, embarrassed, and deeply shaken at his having falling into an abyss of avoidable deception.

Some defend web polls as “participation.” That is bunk. Sheer unadulterated bunk. Web polls are a way of fabricating what popular perceptions really are. They provide only the illusion of participation.

Benjamin Disraeli had the privilege of not living to see web polls. If he had, he might have remarked that there are four kinds of lies, “Lies; damn lies; statistics; and those damnable abominations of mendacity, web polls.”

The great, astounding levels of statistical ignorance that are everywhere abundant is a danger to the continued freedom and prosperity of our modern society. Our educational system must teach every student basic statistics. We adults must continue to educate ourselves. For all that is at stake, we must stamp out statistical insanity.

Update: 20 December 2003. Linked by Walloworld. (‡)

This entry was posted on Thursday, December 18th, 2003 at 23:37 and is filed under Other web sites, Media. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.