skip to primary navigation skip to content

Studying at Cambridge

DAPA Measurement Toolkit

 

Introduction

;

Epidemiological terms are used throughout the toolkit to describe the attributes of methods in terms of quality and feasibility. Many of these terms have scientific meanings that differ slightly from their non-scientific or day-to-day usage. In these sections, concepts which are fundamental to the measurement of diet, physical activity and anthropometry are explained, including:

  • Validity
  • Reliability
  • Error and bias
  • Feasibility
  • Data processing
  • Statistical assessment of reliability and validity
  • Harmonisation

Estimates of diet, physical activity or anthropometry should be as close to the true value as possible. The process of estimating this true value involves several stages of inference. Errors can be introduced at each of these stages, weakening the inference, and resulting in estimated values which do not fully represent the true value. An example of these stages of inference is shown in Figure C.1.1, which is split into two levels:

The data acquisition level, which can be further separated into:

  • Raw measurement
  • On board feature extraction – initial filtering or transformation of information (e.g. brain filtering out unimportant sensory information, or transformation of raw acceleration into physical activity counts)
  • Data storage

The time-independent level, which can be further separated into:

  • Post-processing – removal of invalid data, dealing with outliers etc.
  • Feature extraction – intermediary summary variables derived from the data in order to aid prediction (in some cases extracted features are the final estimates)
  • Prediction – the final estimate of the true value of the behaviour or characteristic of interest which would be reported in research output
Figure C.1.1 Stages of inference involved in predicting the true value by measurement using the tool and overall method.
Adapted from: [2].

Tool

Tool, sometimes ‘instrument’, refers to the mechanism by which data are initially acquired in the data acquisition phase (see Figure C.1.1 above). Tools produce measurements.

There are often different tools which fall under the same category, for example there different examples of food frequency questionnaire, such as the Harvard or Willett questionnaire [3] and the Block questionnaire [1]. Similarly, there are variations in the technical specifications of different accelerometer models used to measure physical activity.

Method

Method is the term given to the combination of the tool used in the data acquisition level, and how the measurements are used to derive estimates during the time-independent level.

Measurements vs. estimates

A tool produces a measurement (e.g. of number of servings of fruit per day), whereas a given method provides an estimate of a target variable of interest (e.g. Vitamin C intake) through a series of inferential steps. In some circumstances, the measurement is used in its raw form (e.g. height measurement), and the estimate and measurement are the same.

Tools as components of different methods

The way in which measurements by tools are used to derive estimations of the target variable of interest varies by study, and sometimes within a study. A single tool can consequently be a component of multiple different methods that produce different target variables.

Additional information as components of methods

Additional information, not captured by the tool itself, can also form part of the overall method. This can include:

  • Participant information such as age, sex, height and weight
  • Energy cost tables for use in conjunction with physical activity questionnaires
  • Nutritional databases for use in conjunction with weighed food diaries
  • Standardised tables for height and weight used to determine growth vs. the general population

Tools vs. brands

Some tools are developed by companies and assigned brand and/or model names. For example, there are many different brands of accelerometer for measuring physical activity, and each of these brands might have various models with different features. Similarly, some anthropometric tools are branded (e.g. Seca), as are some dietary analysis software products.

References

  1. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, Gardner L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol. 1986;124(3):453-69.
  2. Corder K, Ekelund U, Steele RM, Wareham NJ, Brage S. Assessment of physical activity in youth. J Appl Physiol 2008;105(3):977-87.
  3. Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol. 1985;122(1):51-65.