Issues in Survey Design

National data on food availability is generally collected with food balance sheets. While not a survey in the formal sense, this is a collection of data from the food sector regarding wholesale distribution. After adjusting for expected losses and wastage, these data are compared to nutrient values and then to the size and composition of the population to calculate per capita nutrient availability. Because this is a crude assessment, it generally does not account for all losses or waste and therefore tends to overestimate availability.

In order for a household level survey to be nationally representative, one must carefully consider the sampling design. This is generally done by multilevel selection of regions, then sub-regions, then households, in such a way that the resulting data may be generalized to the national level. In cases where results are required at the regional level, coverage of all regions is necessary, although this will usually increase the cost of the survey. When the objective is more specific than national description, target areas may be selected, based on risk status or relevance to the question being addressed.

Similarly, for individual level data to be representative of the greater population of individuals, complex sample design is employed to be sure individuals are selected randomly. Decisions on sampling design will generally be a balance between equal opportunity for subject inclusion against logistic and cost considerations of full randomization. For that reason, multi-level complex sampling design is usually employed. This design may be similar to that of the household level survey, with the added step of randomly selecting individuals within households. Although some surveys do select households and then interview all members of the household, this decreases the generalizability of the individual data due to the lack of independence of the observations. Members in the same family, for example, will consume similar foods and therefore will be more like each other than like others in their community. Although this lack of independence can be adjusted in the analysis design, it will require larger total numbers of interviews to achieve representative stability of data estimates and is therefore not usually the most effective design. While the multistage approach of region, subregion, and community also leads to reduced power, this is corrected by consideration of the 'design effect,' which can be calculated by comparing the variation in intake within versus between sampling units at each level. Although the design effect may demand higher overall numbers of surveyed individuals, this is generally considerably less expensive than expanding coverage to all locations.

In addition to the multistage selection of respondents for representation of the general population, many surveys are concerned with subgroups that will not be well represented unless specifically over-sampled. Examples may include pregnant women, ethnic subgroups, or low-income groups. In these cases, individuals that meet the specified characteristic are identified within the existing sampling design, but are then selected in larger numbers than would be representative of the entire population. This allows sufficient sample size to present valid estimates for these groups. When included in measures of the total population, the extent of over-sampling by subgroups can be adjusted using weights that correct for what would otherwise be an over-representation of these groups.

Another design consideration relevant to accurate representation of dietary intake is the timing of the survey. In many countries, intake may vary considerably by season, and it is therefore important that all seasons are represented. Although logistic and cost constraints often limit ideal design planning, it is also optimal if data from all seasons is collected in all survey locations, as opposed to certain regions being collected in the summer and others in the winter. If the latter is the case, comparisons across regions may be compromised. Similarly, intakes are known to differ by day of the week, and overall intakes may be misrepresented if certain days of the week are not included in the data collection plan.

Keep Your Weight In Check During The Holidays

Keep Your Weight In Check During The Holidays

A time for giving and receiving, getting closer with the ones we love and marking the end of another year and all the eating also. We eat because the food is yummy and plentiful but we don't usually count calories at this time of year. This book will help you do just this.

Get My Free Ebook

Post a comment