Research question. The research question is the abstract question that the research aims to answer by concrete evidence and is the core of any research. It comes at the end of the literature review because to put a question one must define the concepts behind it. It comes before the method as one can’t gather evidence for a question one doesn’t know. The literature review makes the case that it is a good question, given what others have done. A good research question should be:
- Unitary. This means it can be stated in a single sentence, although one research question can break down into many sub-questions, a., b., c…
- Real. “Does fear make one afraid?” isn’t a real question as fear makes one afraid by definition. This is like asking “Is 1 = 1?” A research question puts a choice with more than one possible answer. If the answer is a foregone conclusion, it is not really research. Good research asks risky questions.
- Answerable by evidence. e. g. “What affects life?” is a bad research question as everything affects life and one can’t collect data about everything. A more focused research question makes it more likely that the research will finish.
It can take years to define a good research question, even with advisor help. Often it is trial and error, as one question might be too general to answer while another might have been done before, and so on. Don’t be discouraged if it takes a while, as a weak question makes everything that follows weak. On the other hand, a good question will be good research, even if the data doesn’t show what was expected! After all, knowing what is not true is often just as important as knowing what is, e.g. clinical trials of a drug that doesn’t work. A good research question is a single question that puts a real choice that is answerable by evidence.
State a hypothesis. If possible, state the research question as a hypothesis, a statement whose opposite the research results can deny. The null hypothesis is the opposite of an hypothesis, e. g. the question “Do people prefer polite software?” allows the hypothesis “Subjects use polite software more than impolite software“. This is not “proven” but the null hypothesis that “Subjects will use polite and impolite software the same” can be rejected because assuming that polite and impolite responses are the same, one can estimate statistically if the results agree or not. If the results are very unlikely, one rejects the null hypothesis and so accepts the hypothesis. Argue each hypothesis as a statement whose opposite the results can falsify.
Significance. Significance is the degree to which the null hypothesis is unlikely. It measures the level of confidence in the results. Common confidence levels are:
- Likely. A finding is likely if the probability that the null hypothesis is true is less than 5%. This is 1 chance in 20, stated as p < 0.05.
- Very likely. A finding is very likely if the probability that the null hypothesis is true is less than 1%. This is 1 chance in 100, stated as p < 0.01.
If p < 0.05, the null hypothesis is rejected and the hypothesis is supported, else it is not supported.
Where’s the proof? So where then is the proof in science? In the natural sciences there is none! We don’t live in a perfectly ordered world where 100% proof is possible, as quantum theory confirms that all physical events have some randomness built in. Except for artificial symbolic domains like mathematics, life is uncertain. All science is no more than a set of working hypotheses that are probably true, all open to amendment, e.g. to say that objects fall down by the earth’s gravity just means that we haven’t found an exception, and indeed in theory an anti-matter object on earth would “fall up”. Some find this depressing but for me it makes science interesting. A scientist is one who accepts uncertainty and finds it interesting.