Myra Samuels Memorial Lecture

The Statistical Problem of Relating Nutrient Intake and Disease

Professor Raymond J. Carroll
University Distinguished Professor, Professor of Statistics
Texas A&M University

Start Date and Time: Fri, 16 Apr 1999, 4:30 PM

End Date and Time: Fri, 16 Apr 1999, 5:30 PM

Venue: MATH 175

Refreshments: 5:30 p.m. in the Mathematical Sciences Library Lounge

Abstract:

Most observation studies in nutritional epidemiology attempt to relate disease and nutrient intakes by measuring the latter using an instrument called a food frequency questionnaire (FFQ). The FFQ is a self-report of typical eating patterns and amounts, and is as such subject to biases and errors of measurement. Recently, there has been considerable controversy concerning the existence of a relationship between total fat intake as measured by the FFQ and breast cancer. I will review this controversy and discuss the subtle statistical issues that arise. In particular, the issue of whether the FFQ has any statistical power to detect a fat-breast cancer relationship will be discussed. Analyzing the issue inevitably leads to an investigation of how FFQ's are related to real (usual) intake of a nutrient. I will present a simple statistical model that is sufficiently flexible as to provide a continuum between the extremes of the FFQ being a powerful instrument and the FFQ being of little value, and I will discuss a new study which will attempt to resolve between the extremes.