Accuracy of visual estimates of plate waste in the determination of food consumption.

A six-point visual estimation scale developed by Comstock was used by qualified dietitians to rate amounts of 978 servings of 117 different food items remaining as plate waste. Consumption of the food items was estimated as the difference between the weight of a randomly selected full serving of each item and the score on the visual estimation scale, which had been expressed as a proportion of that full-serving weight. The bias of the visual estimates averaged 2.2 gm; their imprecision averaged 13.7 gm. The small amount of bias in these estimates only slightly influences the results of studies using them as a substitute for weighed data. However, the much larger amount of imprecision must be accounted for. Sample size requirements for the detection of a statistically significant difference in food consumption between two or more groups using visual estimates of plate waste will be greater than those required when exact weighing techniques are used. In addition, the accuracy of dietary self-report data will be underestimated when such visual estimates are used as criterion measures of food consumption unless statistical adjustments are made to correct for the impact of the documented imprecision.