Considerations in Planning Analysis Strategy

Preparing a detailed analysis plan is important in having quality, reproducible research. This page outlines strategies and provides examples to be considered before data analysis begins.
  1. Identify overall goal of the study
  2. Identify specific aims and how they relate to the overall goal
  3. Identify the current state of scientific knowledge
  4. Identify the competing hypotheses that the study is designed to discriminate between, or the quantity of interest that is to be estimated
  5. Refine scientific hypotheses into statistical hypotheses or estimands
  6. Identify type of question
    • Prediction, estimation, or testing
    • Identifying groups, quantifying distributions, or comparing distributions
  7. Where appropriate, specify statistical hypotheses or estimands in terms of a summary measure for the distribution of measurements
    • e.g., mean, median, proportion, event rate
    • Create table shells
  8. Consider design of ideal experiment
    • Ignore practical, ethical limitations in order to be able to later compare how close the actual situation is to the ideal
      • Who would be the subjects
      • What would be the intervention
      • How would subjects be assigned to the intervention
      • What would be the variables measured
      • What are the subgroups
  9. Data Collection
  10. Sampling scheme
    • Retrospective vs prospective
    • Observational vs intervention
    • Inclusion, exclusion criteria
    • Sample size determination
  11. Variables
    • Names
    • Relationship to real world quantities
    • Measurement conditions
    • Units of measurements
    • Categorization according to meaning
      • Demographic, baseline physiology, baseline disease risk factors, prognosis, measures of treatment intervention and outcome
    • Categorization according to use in analysis
      • Response (outcome) variables
      • Predictor variable of interest (variable identifying groups)
      • Variables identifying subgroups to explore effect modification
      • Potential confounders
      • Variables which allow increased precision
      • Surrogates for response
      • Correlated data
      • Missing data
      • Interactions
  12. Type of analysis
  13. Know your audience
When creating an analysis plan for a grant, additional suggestions may be found in Tips for Writing a Statistical Analysis Plan. Acknowledgment of Scott Emerson for much of this outline and Ben Saville for edited, analysis plan examples.

-- MarioDavidson - 28 Dec 2010
Topic attachments
I Attachment Action Size Date Who Comment
Condensed.AP4.pdfpdf Condensed.AP4.pdf manage 133.5 K 30 Dec 2010 - 11:48 MarioDavidson Example Analysis Plan
Condensed.AP5.pdfpdf Condensed.AP5.pdf manage 91.3 K 30 Dec 2010 - 11:48 MarioDavidson Example Analysis Plan
Topic revision: r7 - 05 Apr 2013, MarioDavidson
 

This site is powered by FoswikiCopyright © 2013-2022 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Vanderbilt Biostatistics Wiki? Send feedback