Hypothesis Development

What makes a hypothesis testable?

Two things. It has a prior basis someone else could check, and you can write the sentence describing the world in which it is false, in the same measurement units the study will use. If nobody can say what result would end the hypothesis, it is not testable in practice however formally it is worded.

What is the difference between a hypothesis and a hunch?

The prior basis, and nothing else. A hypothesis rests on named, dated evidence, on a stated mechanism someone could disagree with, or on the organisation's own operational data. A hunch may well turn out to be right, but at the design stage it is a guess that will bias the instrument, so it belongs in an exploratory register rather than in the hypothesis set.

Should every study have hypotheses?

No, and the default answer in commercial research is wrongly yes. Exploratory and discovery work, problem-finding, ethnographic scoping and early concept work are damaged by hypotheses: they narrow the study to what was already imagined and make the most valuable outcome, a finding nobody anticipated, harder to see. State the decision to carry none, and give an area of enquiry and a stopping rule instead.

What does the null hypothesis mean in practice?

It is the specific state of the world your design has to rule out, written in your own measures, not an abstract "nothing happens". Two things follow. Failing to reject the null is not evidence the null is true, and a study too small to detect the difference that would matter cannot produce an interpretable null at all.

Should a hypothesis be directional?

Only where the prior evidence specifies a direction, not where the direction merely seems likely. Practice is genuinely divided on whether a directional prior justifies a one-sided test. The stronger position in applied commercial and policy work is against it, because a large effect in the unexpected direction is often the commercially important result and a one-sided test forfeits the ability to report it.

Why write down the disconfirming evidence before fieldwork?

Because a disconfirmation rule written after the data has arrived is not a rule, it is a rationalisation, and nothing in the wording distinguishes the two. Writing it first also forces the design question that matters: could this study actually produce the result that would prove us wrong? If not, the hypothesis is untestable by this design.

What do I do when the client has already told me what the study will find?

Record it as a belief with the holder named, which is a fact about the organisation rather than about the world. Then decide whether it can be converted: it needs an evidential prior rather than a positional one, a disconfirmation rule the sponsor agrees to in writing before fieldwork, and a stated commitment to report the result whichever way it comes out. A proposition that cannot survive those three conditions is not a hypothesis.

Is analysis discovered after seeing the data invalid?

No. It is exploratory, and exploratory analysis is legitimate and often the most interesting part of a study. The problem is reporting it as though it had been predicted. Keep a dated register of what was specified in advance, and label everything else as exploratory, with the number of analyses run disclosed.