Study Design Exercises 

1/19/07                   Reference: Some Aspects of Study Design by Gerard E. Dallal, Ph.D.


  1. What is the distinction between experimental and non-experimental designs?
  2. What does double-blinded, randomized, controlled" mean?
  3. If statistics is not merely a compilation of computations methods, what then is it?
  4. Identify a problem associated with the use of "self-controls."
  5. Compare and contrast intention-to-treat analysis with efficacy analysis.
  6. Complete this sentence: According to the article by Dallal, the four "basics of study design" are . . .
  7. What is the difference between a longitudinal measurement and cross-sectional measurement?
  8. What is the primary benefit of randomization?
  9. According to Dallal, what are two concerns researchers must consider before conducting a meta-analysis? (Emalie Huriaux)
  10. When would an observational study be conducted rather than an experimental study? Provide example of each. (Janet Hughes)
  11. Why would you use an active control as opposed to a placebo control? 


SD.1 Measurement error vs. processing error. We may distinguish between measurement error and processing error. Measurement error (information bias) is the difference between the true value of an observation and what appears on a data collection form. Processing error occurs after data have been collected. Of the two, which presents the greater hazard? Explain your reasoning..

SD.2 We can't we experiment?  Most of Dallal's article discusses the benefits of experimentation. However, many human problems cannot be studied experimentally. Provide an example of a human health problem that precludes experimentation. Identify the specific ethical dilemma involved. How would you study this problem non-experimentally? 

SD.3 Design a study protocol. The goal of this exercise is to design and document a modest study. Your project should be well-focused and include  measurements on about 5 to 10 variables.In designing your study, write a protocol that addresses each of the following design components: 

(A) Research Question and Hypothesis: What problem motivates the study, and what question is being addressed? What is research and statistical hypotheses do you wish to address? What parameters are being estimated? 

(B) Target population: What person, place, and time factors define the target population? How will this population be sampled?

(C) Explanatory variable (i.e., the independent variable; the exposure variable in an epidemiologic study): What is the main independent (explanatory) variable in your study? How will this variable be defined and measured? 

(D) Response variable (i.e., the dependent variable; the disease variable in an epidemiologic study): How will the study's outcome be defined? Will data be self-reported or independently measured? If self-reported, will findings be confirmed independently?

(E) Extraneous variable (i.e., potential confounders). Will extraneous factors that might confound your analysis be considered? If so, how will they be dealt with?

(F) Anatomy of the design: Is the study observational or experimental? If observational, is it cohort, case-control, or cross-sectional? Will the timing of data collection be concurrent, historical, or mixed? Will study subjects be selected based on their exposure-status, disease-status, or neither? If the study is experimental, describe the randomization method you will use, while discussing blinding methods and ethical considerations.

(G) Planned analyses: Which descriptive and inferential techniques will be used? What summary statistics will be reported? Will graphs be included in your final report? What parameter(s) will be inferred? Will inference be in the form of confidence intervals or hypothesis tests (or both)?

(H) Sample size requirements: How many people need to be studied? Complete preliminary sample size requirement calculations, while being explicit about your assumptions.

(I) Study limitations: Consider possible sources of error. How will you guard against selection bias, information bias, and confounding? List limits to generalizability.

Provide a concise write-up of your protocol. The write-up should be double spaced, and need not be more than a couple of pages long. Please label each section of the write-up with the headings provided above. .Include skeletal computer files you will use during your investigation. If you are working with EpiData, include your QES file, an empty REC file, and a data documentation (DD) file showing variable names and types. [Click here for a student example]


Key to Review Questions and Odd Numbered Problems                                  Key to Even Numbered Problems (may not be posted)