The null hypothesis is not a logical fallacy. We are discussing it as a base for further discussions of fallacies in future posts on the subject of the risks of action vs. inaction and Type 1 and Type 2 errors.
Outside of the world of statistics, the "null hypothesis" has become equated with the "nil hypothesis," which means, basically, nothing. That is to say, that nothing occurred that was not by chance or accident, or maybe by undiscovered or undiscussed causes. Thus it is a handy tool to use as a starting point for an honest discussion, debate, or argument.
The null hypothesis is what many logical arguments ultimately argue about, or around, whether it is made explicit or not (it is a basic assumption of thinking in Western Civilization). When a null hypothesis is not assumed, a case for something is often termed "biased." (As we will discuss in a future post, "bias" is often a very useful and reality-oriented posture, and is the reason we do not look for Bluebirds in Brooklyn.)
In law, the null hypothesis is the presumption of innocence. In science, it is the presumption that there is no connection between two phenomena. (Scientists and social scientists often complain that it is difficult to publish papers which support null hypotheses.) Hypotheses other than null hypotheses are often termed "alternative hypotheses." In general, it is easier to destroy an hypothesis than to prove one: proof is usually too much to ask for.
Let's take one incendiary example:
Null hypothesis: Blacks are not economically discriminated against, and there is nothing of interest here to debate or discuss.
Fact: Black households have lower incomes than white households.
Hypothesis #1: Employers pay blacks less money, or blacks get lower-paying jobs because of their color.
Fact: Black households with intact marriages have essentially the same average family incomes as whites, but blacks have very high rates of unmarried families.
Logical conclusion: An extraneous factor, such as marital status, may be determining the data, not skin color. The null hypothesis is supported by these facts.
A liability example:
Null hypothesis: Jim is innocent of liability or neglect.
Fact: Jim had no proper fence around his pool, and the neighbor's beloved Shitsu wandered over, fell in, and drowned, so the neighbor wants $100,000. for pain and suffering.
Hypothesis: Jim is guilty of not properly fencing his pool.
Fact: Hurricane Jose knocked down his pool fence a week ago.
Logical conclusion: Facts support the null hypothesis. Jim is innocent of negligence because of an accident of nature.
A vegetable example:
Null hypothesis: What you eat has no relationship with colon cancer.
Fact: People who eat lots of broccoli have lower rates of colon cancer.
Hypothesis #2: Broccoli helps prevent colon cancer.
Fact: People who eat broccoli tend to eat lots of other veggies too.
Hypothesis #3: Eating lots of veggies helps reduce colon cancer rates.
Fact: Volume of dietary roughage (cellulose) probably correlates with reduced rates of colon cancer.
Logical conclusion: The null hypothesis is probably wrong. There is some relationship, although causality is not demonstrated (that would be a cum hoc ergo propter hoc fallacy - a favorite fallacy of litigators). You might reduce your risk of colon cancer a bit by eating plenty of daily veggies and salads. (Still, your genes - or your GI doctor - may determine the outcome, eventually.)
We wll build on this null hypothesis subject in the next Fallacy posting, which will highlight Type l and Type ll errors.
More on the subject of the very important Null Hypothesis here. (I enjoy giving myself this elementary refresher - hope you like it too. Next installment probably on Thursday.)
