From a food consumption survey to an estimate that supports a food and nutrition decision
Population dietary intake, intervention studies and impact evaluation of food programmes, handled the way self-reported data require.
Nutrition data are largely indirect. A 24-hour recall describes what was consumed on one day, not a person's usual intake; a household survey records shared foods, not how much each member received; and knowing the iron or zinc content of a food is not the same as knowing how much the body can absorb. Each of those gaps introduces a different source of uncertainty.
We work with nutrition research teams, foundations and food programmes with a population component. The statistical work consists of making that chain explicit: what was actually observed, which part has to be estimated, which assumptions allow you to move from one source to another and how far the conclusion can be taken. The design of the intervention and knowledge of the ground stay with the team; we build the estimate, model the scenarios or effects the design allows and quantify their uncertainty.
The project with New Seed in Guatemala is the backbone of this sector: several years of work on biofortified maize that connect intake estimation, producer characterisation, adoption modelling and evaluation of the potential nutritional impact.
We regularly work with
Nutrition research teams working with food consumption surveys who need to move from reported foods to a nutrient estimate that respects how the data were collected.
Foundations and organisations that run food programmes and need to separate what was observed from what was modelled when they report on coverage, intake or impact.
Groups that have designed a dietary intervention and need to justify its size, define its outcomes and analyse change without losing the longitudinal structure of the study.
Projects with several waves, instruments or sources that do not fit together directly and need harmonising before any comparison between groups or periods makes sense.
Where we usually come into a nutrition project
In nutrition, the entry point is usually set by where the uncertainty appears. It may sit before the study, when designing an intervention or an instrument; in data already collected, when there are repeated measures or biological sources; or in the intake estimate itself, when reported consumption has to be translated into nutrients per person and subgroup. If you would rather start from what you need to do than from your sector, each block leads to the full path for that need.
Design a study
A study is won or lost before the first data point is collected.
Design and analysis of studies that measure the effect of a dietary intervention on health
View service →Validate the instrument and analyse what it measures: reliability, clinical scales and patient-reported outcomes
View service →I have data and don't know where to start
You already have the data. The question is what to do with them.
From repeated measurements over time to an understanding of how things change
View service →From sequencing data to integration with the rest of your study
View service →How patients move between states, where their trajectory changes and where follow-up is lost
View service →Monitor populations and places
Understand what is happening to a population before it becomes a headline.
The rest of the catalogue, organised by what you need to do
A nutrition question can end up needing tools that are not specific to nutrition: causal inference to attribute a change, simulation to anticipate a scenario, meta-analysis to review the existing evidence, or database preparation to integrate incompatible sources. Those services are organised by the need they address, not by the discipline in which the question is framed.
Projects in this sector
What we are usually asked in this sector
Our consumption data are self-reported by participants. Is that an obstacle for the analysis?
No, but it means the data have to be analysed for what they are. A recall or a food consumption questionnaire contains measurement error and, depending on the instrument, a mix of usual variability and the variation of a particular day. The analysis has to distinguish which part can be corrected or modelled and which part will remain part of the uncertainty. When the design allows, usual intake can be estimated, subgroups described and prevalences of inadequacy calculated; what should be avoided is treating a one-off report as if it were a direct, exact measurement of the diet.
How is the impact of a food programme on a population's nutritional status estimated?
It depends on whether the programme has already been observed or is still being planned. If it is already in place and the design allows a valid comparison to be built, its effect can be estimated against a counterfactual, with the assumptions behind that attribution made explicit. If it has not yet been implemented, what you get is an ex ante projection: a model of how a change in coverage, consumption or composition could carry through to the indicator of interest under given assumptions. Both questions are useful, but they do not mean the same thing and should not be presented with the same verb.
We have data from several waves and several instruments. Can they be analysed together?
Sometimes, but comparability has to be demonstrated before anything is combined. First we check whether the variables keep the same definition, unit, reference population and method of measurement across waves, and whether the instruments really capture the same construct. What can be harmonised can be integrated and analysed longitudinally; what changed meaning has to be transformed with an explicit justification or kept separate. The aim is not a single database at any cost, but comparisons over time that keep a defensible meaning.
Does your work fit better elsewhere?
Shall we talk?
Tell us where your programme or study stands and we'll give you a straight answer on how we can help.