For some questionnaires, responses may be incomplete and therefore coded as missing values. This is most likely to be a problem in surveys where a paper form is used and the questionnaire is self-administered. If the questionnaire is to be used to estimate nutrient intakes, instructions must be standardised as to whether intake of a particular nutrient should still be calculated when some responses are missing. This will be dependent upon the proportion of the total nutrient intake provided by the missing item in the population being studied. For example, sugar intake should not be calculated for a particular individual if it was not known whether sugar is added to tea or coffee. However, if this were the only missing data on that subject's questionnaire, intakes of other nutrients such as fibre could still be calculated.

You also need to consider how missing values are to be processed. For example, if the size of slice for bread is missing, should bread be omitted from the nutrient intake calculation? Or, should nutrition analysis not be done at all for that questionnaire? Or, is it possible to make an assumption about the size of slice if the frequency and number of slices are known? Each option may have a detrimental effect on the overall quality of the survey data, depending upon which data are missing and the number of questionnaires affected. If the missing items are omitted from the nutrition analysis, nutrient intakes will be underestimated. If the whole questionnaire is omitted from the analysis then the response rate of the survey is reduced and the remaining subjects may not be representative of the population from which they were selected. If assumptions made about missing values are inappropriate for the population being studied, calculated mean intakes may still not be an adequate estimate for that population. The best option for each missing value on the questionnaire therefore needs to be given very careful consideration.