Statistical Report Review (Draft)

Beginning May 2009, the Department of Biostatistics instituted a periodic constructive review of statistical reports prepared for collaborators by faculty and staff biostatisticians in the department.

Purposes of Reviews

• Having non-supervisors who may be expert in certain areas (e.g., statistical programming) review your work
• Continually improving technical writing
• Improving reproducible research practice and graphics
• Helping all department biostatisticians continually improve their analytic and other skills
• Rewarding outstanding reports by prizes, having the statistician present the report at a seminar or clinic, or (with collaborator permission) having parts of their reports displayed on the hallway monitor
• Helping the Department collect an archive of outstanding reports to be shared with department biostatisticians for their continuing education

Areas Covered in Reviews

• Technical writing
• Quality summaries and references that a collaborator can insert into a grant proposal or manuscript
• Appropriateness of statistical analysis
• Quality of explanations of analysis methods and interpretation of results to collaborators
• Quality of statistical graphics (informative, accurate, free of optical distortions, little need for remote legends; key graphics publication ready)
• Programming efficiency from a person-time perspective
• The use of good programming practices as described in ProgrammingTipsForStatisticians (for example, not using code that depends on the data or category levels being in a certain order)

Reports should be submitted in a format so that commands that produced each calculation, graphic, or table accompany that output. Users of Sweave will have an advantage in this regard, when they use Sweave defaults that shows all of the code.

Frequency of Reviews and Feedback to Supervisors

Each statistician producing statistical reports should submit a report for review quarterly. There will be a few exceptions when the statistician's supervisor already provides thorough reviews. Reviewers will seldom be supervisors. The reviewer will provide comments to the statistician by marking on the report, and will provide a general impression to the supervisor (e.g., "above average", "average", "improvements needed").

Review Committee

The Chair will assemble a committee of approximately 10 individuals who are competent in statistical computing. Reviews will be assigned sequentially, cycling through the list to give a maximum break between reviews. The department will develop a reward for completing reviews, such as some limited travel funds.

The committee members and dates they have completed reviews are found below. Please let Frank know if you are interested in being a reviewer.