%Title: On Statistical Significance %Author: Beracah Yankama %AuthorTitle: Director, StudentsReview When students and parents are reading reviews, often-times the question will come up, "how statistically significant is this data?". How can only a few reviews be representative of several thousand students' opinions? Few people really understand how statistical significance works or what it really means. While statistical significance itself has a meaning rooted in replicability of the data, most people interpret it as, "is this enough data to trust?" % put in actual definition here. then next paragraph talk about how people interpret it. % school distribution cannot possibly be replicated; it is changing year to year, semester to semester. Consumers tend to interpret statistical significance as the "quantity needed to dismiss exceptions" It is often used as a marketing term by pharmaceuticals. Consider this: "1/1000 experienced side effects when taking our drug" 'Statistical significance' as many people understand it, will lead you to believe that the 1 person was some kind of an exception, a chance occurrance, the side effects are not likely to happen to you. % (and to the 99.9%), that number leads you to believe that the 1 person was some kind of an exception, a chance occurrance, that % death is not likely to happen to {\em you}. But suppose the test on the drug was given to 999 adults, and 1 child. The % child died. Now is it statistically significant? Statistical significance does not mean anything -- it is only a tool used % to get the consumer, whether drug purchasers, or college applicants, to dismiss the error (and shell out the money) without % knowing more. Suppose, however, that the drug in question was given to 999 adults, and 1 child. The child experienced the side effects. Now is it statistically significant? Invariably about 66% of people are satisfied with their college. 33% however, are not, and wished they had gone elsewhere. The larger percentage would convince you that the 66%, the majority, is all you need -- "most people worked out ok". Because that lets the blame fall on the consumer, not the school. Lets consider one last example. Suppose that a school consists of 95 males and 5 females. The 95 males say the safety of the school is great, then 4 of the females say it is "ok". The idea of Stastical significance might have you (the generic, genderless consumer) believe that the safety is great. You the parent, you the prospective female student, should be rightly curious of any exception. What you {\em really} want to hear is from the 1 girl who might have been attacked at night, who tells her story. Does the fact that 4 out of 5 women said that safety is "ok" matter? Does the opinion of the 95 other men matter? As the parent, the prospective student, you want to know about what happened to that 1 girl, because that tells you that 'your little girl' might not be safe on the campus. You don't want to know "the probability of attack", you want to know a clear-cut, answer to the question, "does it happen, or not?" After all, most people don't die when they cross the street -- but does that mean you shouldn't look both ways? % Lets consider one last example. Suppose that a school consists of 95 male and 5 female. The 95 males say the safety % of the school is great, then 4 of the females say it is "ok". SS would have you (the generic, genderless consumer) % believe that the safety is great. But you the parent, what you {\em really} want to hear is from the 1 girl who was % raped at night, who tells her story. Does the fact that 4 out of 5 women said it is "ok" matter? Does the opinion % of the 95 other men matter? NO. As the parent, you want to know about what happened to that 1 girl, because that % tells you that YOUR LITTLE GIRL might not be so safe on the campus. You don't want to know "the probability of % attack", you want to know "YES, or NO". The "majority opinion" has widespread use to cover up failures, because % it is simply impossible for bad things to happen to *most* people. Most people don't die when they cross the street either % -- does that mean you shouldn't look both ways? StudentsReview is intended to catch the *exceptions*, the people who were misled by marketing, by well publicized majority opinion, into making an incorrect decision. Read the opinions knowing that they are experiences. They *did* happen. They didn't have some "probability of happening", and weren't some kind of "just a complainer" event. Where something happens once, it happens twice, so read, learn, and make an informed decision. Don't let others dupe you into dismissing evidence so that you make the same uninformed decision that they did. Many people simply do not want to face the fact that they may have, or are making a mistake. % StudentsReview is intended to collect the kind of opinion you might get from talking with an acquaintance or close % friend who attended the school. "Statistical Significance" is a marketing lie used to push dangerous drugs on % people, and bad decisions -- StudentsReview exposes the experience behind the student, so that the reader can determine what questions they should be asking for themselves. As always, caution is the best advice, and we've provided a mechanism for students to be cautioned. % 3. StudentsReview is *not* biased positively or negatively. It merely turns out that students in the middle of % the opinion spectrum simply do not have an opinion. Its not because we "attract" those people. In the beginning % that might have been true, but we've grown large enough to attract all kinds of students.