New Approaches to Graduate Biostatistics Training

Biostatistical Rounds for Graduate Students

The ultimate purpose of statistics in medicine is to help clinical scientists to accurately convey the meaning of their experiments to their peers. Even the most sophisticated statistical analysis is of limited utility if the clinical investigator lacks some degree of intuitive understanding or what it means. Moreover, clinicians often come to biostatisticians with somewhat ill-formed notions of exactly which hypotheses should be tested and contrasted, or how their experiments should be designed and analyzed. Major parts of the art of biostatistics consist of learning how to do the following:
  • communicate with clinical scientists,
  • learn enough of the biology to be able to identify likely confounding variables,
  • explain study designs and why specific design components are needed for specific studies,
  • explain the meaning of statistical analyses and to clinical colleagues,
  • assist in the writing of manuscripts that clearly explain how experiments were performed and analyzed and what the findings mean.

In order to build these skills all graduate students in biostatistics will be required to take a course entitled Biostatistical Rounds. Each student will be assigned to a faculty member and will participate in all of her consulting activities over a semester. Each student will be required to complete two rounds with two different professors in order to observe different styles of interacting with investigators. Students will start by attending meetings with investigators and conducting supervised analyses. As the semester progresses they will play a more active role in interacting with investigators and explaining their analyses. Students will be required to be a coauthor on at least one paper and to be actively involved in writing up the methods and results section of this paper. Grades in this course will be determined by evaluations by both the biostatistics faculty member and the clinical/translational faculty who interact with the student. Biomedical researchers will receive free work from the students in exchange for their participation in this teaching activity. During the two semesters of biostatistical rounds each student will identify a clinical, translational, or basic science investigator who is willing to have a more long-term interaction with the student. This investigator will act as a mentor for the student's Biomedical Science Rotation described below.

Biomedical Science Rotation for Graduate Students in Biostatistics

Graduate students in biostatistics need an appreciation of the complexities of modern biologic science and how laboratory technique can affect stochastic elements of medical experiments. We will require that all students be exposed to at least one area of laboratory science. Students will take a four credit-hour course in which they will work as an intern in an investigators laboratory and will read texts or papers assigned by the investigator that will introduce them to the techniques that are involved. The depth of this course will depend on the background of the student, but at a minimum she will observe how experiments are done and will assist in some aspects of this work under careful supervision. Doctoral students whose thesis will involve the analysis of a specific type of biologic lab data will be required to complete their biomedical science rotation in a relevant laboratory and may be required to pursue this work at greater depth than the typical student. For example, a student whose doctoral thesis will involve the analysis of mass spectrometry data will meet their biomedical science rotation in a mass spectrometry laboratory.

Train Biostatistics Graduate Students in Measurement Evaluation

A common theme in laboratory, translational, and clinical research is the use of newly emergent measurement methods to interrogate the biological processes under study. Advances in measurement technology allow unprecedented ability to investigate molecular, cellular, immunological and other processes, with huge potential impact on disease prevention, diagnosis, and treatment. Despite these advantages, new measurement methods frequently exhibit substantial measurement variation, and exhibit poor repeatability and/or reproducibility. Further neither laboratory scientists (e.g., analytical chemists, clinical pharmacologists) nor biostatisticians are typically trained in statistical metrology (science of measurement) for assay development, evaluation, or quality control. We will develop a course for graduate students in biostatistics and for quantitatively oriented biomedical scientists in the statistical development of measures (statistical metrology). This course will introduce to students the following topics:
  • consequences of measurement variation
  • assessment of variation for direct and indirect assays
  • experimental designs to identify sources of variability, and for assay improvement
  • application of statistical quality control to measurement processes
  • normalization methods in high-dimensional "Omics" data
Although such a course would be new in biostatistics and biomedicine, the ideas and basic methodology are well established in the physical sciences and quality industry. A recent book by Hand1 as well as established texts by Mandel2 and Rabinovich3 would serve as useful textbook/reference material.


[1] D. J. Hand. Measurement Theory and Practice. Hodder Arnold, 2004.
[2] J. Mandel. The Statistical Analysis of Experimental Data. Dover Publications, 1984.
[3] S. G. Rabinovich. Measurement Errors and Uncertainties, 2nd ed. Springer-Verlag, 2000.

Topic revision: r2 - 17 Jan 2008, FrankHarrell
 

This site is powered by FoswikiCopyright © 2013-2017 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