From question to hypothesis
Chapter 3: Develop Research Ideas and Questions — Video 5
https://ghrbook.com/videos/from-question-to-hypothesis/

[Slide 1] You have a research problem and you know your learning goal. This video turns those into a question you can answer.
[Slide 2] A good research question addresses a research problem, a gap in our knowledge, and is answerable. Answerable does not mean easy to answer, just possible to answer.
[Slide 3] The question, do mosquitoes have an afterlife, represents a gap in our knowledge, but we don't have any empirical means for finding an answer. It's an interesting question to pose to my kids to probe their imagination, but it's a poor research question, because it's not answerable today.
[Slide 4] Writing good research questions takes practice. Here are two acronyms to help you get started: FINER and PICO. FINER stands for Feasible, Interesting, Novel, Ethical, and Relevant. Use this checklist to evaluate your research questions.
[Slide 5] Feasible. Some research questions will take a long time to answer, cost too much, require too many participants, demand skills or equipment that you do not have, or will be too complex to implement. A good research question is not just answerable, it's feasible for YOU to answer it with the resources currently at your disposal.
[Slide 6] Interesting. Research requires funding and effort. Without an interesting question, you won't get funding. Without sustained interest in answering the question, you might not finish the work, and global health research tends to have long timelines. Novel. Replication is an important part of science, but the majority of global health funding goes to research that asks new and interesting questions, so you should prioritize answering questions that fill a gap in our knowledge.
[Slide 7] Ethical. Google's motto used to be, don't be evil. This is a minimum bar for research with human subjects. You have a responsibility to ensure that the questions you ask and the methods you employ in search of answers shield participants from harm. And Relevant. In addition to being interesting, a research question should be relevant to science and society. The answer should move your field forward. Making that determination requires a thorough review of the literature and conversations with senior colleagues.
[Slide 8] The second acronym is PICO. It stands for Population, Intervention, Comparison, and Outcome. Let's use it to develop a research question about the efficacy of mosquito bed nets in preventing malaria.
[Slide 9] P. Start by defining the target population, or the problem that needs solving. I like to include both. For instance, we might ask a question about malaria infections, the problem, among children under 5 years of age living around the Lake Victoria basin in Kenya, the population.
[Slide 10] I. This can refer to an intervention, an exposure, or a prognostic factor. An example of an intervention for preventing malaria infection is the provision of insecticide-treated bed nets. If you're not conducting an intervention trial, you might instead be interested in an exposure that increases the risk of an outcome, such as traveling to a malaria endemic region. Or a prognostic factor that predicts mortality, such as neurological dysfunction in severe cases of malaria.
[Slide 11] C. The choice of a comparator is a critical aspect of your research question. Are you interested in comparing bed nets to another intervention, such as indoor spraying? Comparing insecticide treated bed nets to untreated bed nets? Bed nets compared to nothing? Or maybe your research question is descriptive and has no comparator. For this example, let's say that we're interested in the effect of treated bed nets on the outcome compared to untreated nets.
[Slide 12] O. Outcomes are the specific targets of our investigation. For instance, we might be interested in estimating the impact of insecticide treated bed nets on parasitaemia, the presence of malaria parasites in the blood.
[Slide 13] We can combine these elements into a single research question. Among children under 5 years of age living around the Lake Victoria basin in Kenya, are insecticide-treated mosquito nets more effective than untreated nets at preventing parasitaemia? PICO works beautifully for intervention questions, but not every research question fits this mold, and the chapter has alternative frameworks for other types of questions.
[Slide 14] Once you have a research question, you might be tempted to jump straight to hypothesis testing. But let's slow down and think about what a hypothesis actually is, and whether you need one. A hypothesis is a tentative explanation or prediction that can be tested with data. In its simplest form, a hypothesis states what you expect to find. For example: insecticide-treated bed nets reduce malaria infection rates in children under five compared to untreated nets. That's a testable claim. You could design a study to gather evidence for or against it.
[Slide 15] But here's something that surprises many students. Not all research requires a hypothesis. Descriptive studies often don't have one; they aim to characterize what exists, not to test a prediction. Qualitative research frequently generates hypotheses rather than testing them. And exploratory work, by definition, is looking for patterns without preconceived expectations.
[Slide 16] So when do you need a hypothesis? Generally, when your goal is causal inference, when you're testing a specific prediction derived from theory, or when you want to examine an association. If you're asking, does X cause Y? Is treatment A better than treatment B? Is exposure C associated with outcome D? Then you should have a hypothesis.
[Slide 17] A hypothesis isn't just any prediction. It must be falsifiable. This means it must be possible, at least in principle, to find evidence that would prove it wrong. Consider two statements. Insecticide-treated bed nets reduce malaria infections. And, bed nets are helpful for malaria prevention. The first is falsifiable. We can design a study, collect data, and potentially find that treated nets don't reduce infections, or that they do. Either way, the evidence speaks to the claim. The second is too vague to test: helpful how? Reduce infections? Reduce severe disease? Reduce vector bites? Under what conditions? Over what timeframe? So when you develop a hypothesis, ask yourself: what evidence would convince me I'm wrong? If you can't answer that question, you don't have a scientific hypothesis.
[Slide 18] Good hypotheses don't emerge from thin air. They're grounded in theory, an explanation of how and why things work the way they do. Theory tells you what to expect and why. Consider the bed net example. Why would we expect insecticide-treated nets to reduce malaria more than untreated nets? Because malaria is transmitted by mosquitoes, insecticides kill mosquitoes, and dead mosquitoes can't transmit disease. This causal chain gives us reason to believe the hypothesis is plausible. It also helps us understand what we're really testing: not just whether treated nets work, but whether the mechanism we've proposed produces the expected outcome. So when you develop a hypothesis, ask yourself: what's the theory behind this prediction? If you can't articulate one, you might be fishing for results rather than testing a meaningful claim.
[Slide 19] Remember that terrifying moment when a mentor asks, what are your ideas? You now have a framework for answering that question. No need to hyperventilate. Good ideas rarely arrive as eureka moments; they emerge slowly, through exposure and connection, as you explore the adjacent possible. You now have the building blocks for designing a study: a research problem, a learning goal, a research question, and, when appropriate, a hypothesis grounded in theory. But a well-crafted question is only the beginning. In the chapters that follow, we'll turn these ideas into action, searching the literature, designing studies, collecting data, and making sense of what we find.