Skip to main content
ARS Home » Research » Publications at this Location » Publication #161905
Title: DOES THE MEAN MEAN WHAT WE THINK IT MEANS IN NUTRIENT ANALYSIS?

Author
item GOSSETT, JEFFREY - DELTA NIRI
item PARKER, JAMES - DELTA NIRI
item SIMPSON, PIPPA - DELTA NIRI

Submitted to: International Biometric Society Conference
Publication Type: Abstract Only
Publication Acceptance Date: 12/1/2000
Publication Date: 3/25/2001
Citation: GOSSETT, J.M., PARKER, J.G., SIMPSON, P.M. 2001. DOES THE MEAN MEAN WHAT WE THINK IT MEANS IN NUTRIENT ANALYSIS [abstract]? Proceedings, International Biometric Society Eastern North American Regional Conference. p. 120.

Interpretive Summary:

Technical Abstract: Introduction: In most nationwide nutritional studies, data analysis is complicated by the use of complex survey designs. Reported summaries are typically weighted means with standard errors. The mean and standard error are not sufficient or appropriate statistics for non-normally distributed data. We investigate whether reporting means in nutrient data, which are often skewed, is misleading. Methods: Data from the 1994-1996 CSFII survey is used. This survey was designed to obtain nationally representative samples of households in the United States. We use adults, male and female, aged 20 to 44 and 45 to 64 giving males 20-44 (n=2127), males 45-64 (n=1733), females 20-44 (n=2023), and females 45-64 (n=1674) for a total of 7557 people. We examine the nutritional endpoints caffeine, calcium, carbohydrates, energy (calories), protein, and total fat. All analysis is done using SUDAAN. Results: The medians consistently were smaller than the lower 95% confidence interval (CI) of the mean. In 20 of the 24 groups, the upper 95% CI for the median was less than the lower 95% CI for the mean. Summary: Reporting the mean gives a falsely high impression of consumption.