CLIO Learning Modules
Study DesignSelectionSample SizeData Collection and AnalysisHuman Subjects

'Me too studies'
Target Population
Attributable Risk
Relative Risk
Data Sources
Study Time
Case Control
Nested Case-Control
Prospective Cohort
Retrospective Cohort
Randomized Clinical Trial
What is your Hypothesis?

  • Understand the role of hypothesis formulation in a research plan.
  • Recognize the main elements of an hypothesis.
  • Apply criteria for formulating an hypothesis.
  • Write your main hypothesis (pick the most important).
  • Formulate the null hypothesis (proposition to be disproved).


The hypothesis is a conjecture that you intend to test with the results of your study.


The first step of designing your study should be to develop a clear hypothesis - it will be the central focus during the design of your study. Every decision regarding study design should ultimately return to the question "How will this affect my ability to test my hypothesis?"


Hypothesis: "Palm Pilots are bad."

This may be in fact what you believe and want to prove. However, it makes a poor hypothesis for a study, because it is vague. What is bad? Bad for whom?

Improved hypothesis: "Palm Pilots cause brain rot in medical students."

While the sentiment of the original hypothesis is still present (assuming of course, that you consider brain rot in medical students to be bad), it is vastly improved because it clearly states the proposed association between an exposure* (Palm Pilots) and outcome* (brain rot) in a target population* (medical students). *These topics are covered in detail in subsequent modules.

Knockout hypothesis: "Daily use of Palm Pilots by medical students results in atrophy of the hippocampus."

Wow. The exposure, outcome and target population are crystal clear. The added details in the exposure and outcome have also made it more testable. With a hypothesis like this, your study practically designs itself - it will involve a group of medical students, assessment of Palm Pilot usage, measurements of their hippocampi, and some statistical test of the association between usage and atrophy.


As the above example illustrates, having a clear, well-considered hypothesis makes study design a snap. All that's left is to collect the data, analyze it, and send it off the New England Journal of Medicine. Reality, of course, is a bit more complicated. During the course of conducting a study or at the time of data analysis, you may find yourself pursuing questions that are unrelated to your hypothesis. Be very cautious. It is perfectly legitimate to have secondary hypotheses, but they should be formulated before conducting your study. Remember that your study was designed to test your hypothesis - exploration of tangential issues may be subject to biases that your original hypothesis was not. Generate your study hypothesis with great care and consideration; revise and adapt as your study ideas crystallize and before your study begins.

June 4, 2004 v0.20
Copyright © 2004 Stanford School of Medicine