Scientific studies have often provided useful information which have led to positive and beneficial courses of action. Some studies are less useful than others because there are ways in which they can be flawed. This section is an introduction to how studies and statistics work and why they are sometimes useless.
The population of Florida is increasing and parts of Florida are sinking. Therefore the extra weight of extra people is causing Florida to sink. This is obviously false, but it illustrates how 2 things happening simultaneously are not necessarily related.
Married men live on average than unmarried men. Therefore if a man marries he increases his life expectancy. This is a false assumption because men who never marry and married men are different populations. Men who never get married include many subgroups with a lower than average life expectancy. For example:
A typical study would involve a hypothesis say "circumcision protects against aids". The use of a hypothesis is a powerful idea, though the downside is that the experimenter may have an aggenda of proving his idea, rather than searching for truth. Typically, 2 groups of subjects are found. In this case it would be intact men and cut men. The incidence of HIV in each group would be compared.
Different statistical tests have been invented to see if a difference (if any) is unlikely to be due to chance. For this to happen there usually tends to be quite a difference.
Samples cannot be too small. For example if I had a hypothesis that smoking made teenagers thin and I had 3 teenagers in my study, the sample would be far too small.
The populations of the groups being compared need to be as similar as possible (except for the difference in the hypothesis). In practice this can be very difficult.
Also the sample populations should be a true sample of the overall population so that the conclusion can be generalised to a larger group, otherwise a study will not have much use.
Many studies have a control group of subjects. This is a group not subjected to an experiment, against which the subjects can be compared.
Some studies work best if they are double-blind, that is the experimenters don't know who is in a control group and who is not.
Many studies fail because of faulty sampling. Years ago, a telephone poll in which voters wre asked who they would vote for failed to predict the election winner, because at that time, many poorer people had no telephones. Telephone owners were not representative of the total population.
Getting proper samples can be very difficult. Subjects may be volunteers and not representative of the general poulation. Or they may be people who tend a VD clinic. A study may be done very poorly.
For example, comparing cut and intact men, might not take into account that the groups are different in many ways. In some countries cut men in a study might be entirely Muslim, while intact men might be entirely non-Muslim. The non-Muslims might have more HIV, but this could be due to any number of reasons, for example monogamy versus promiscuity.
One circumcision study involved comparing cut Filipionos with intact Spaniards. This is not a proper study.
One study in an army hospital showed that intact infants had a lot more UTI than cut infants. What the study didn't say was that army doctors typically forcibly retracted foreskins, causing UTI in intact infants. Later studies didn't repeat the results.
Unfortunately, the flawed study by Wiswell is mistakenly quoted as evidence for circumcising. Studies need to be questioned more.
Some of the best discoveries are serendipity, finding something accidentally when looking for something else.