## Glossary of Terms

0-9

A

B

C

D

E

F

G

H

I

J

K

L

M

N

O

P

Q

R

S

T

U

V

W

X

Y

Z

### 0-9

### A

- Absolute Validity
- The extent to which a measure exactly captures the concept it intended to reflect. This is usually assessed in comparison to a 'gold standard' measure.
- Accuracy
- The extent to which a measured value is close to that of the true value. Low accuracy can be a result of bias or systematic error in measurement.
- Actual Nutrient Intake
- Reflects nutrient intakes over a specified period of time, for example from a 24h recall or food diary. Taking into account cooking and food portion sizes. This is in comparison to usual nutrient intake which would be obtained from an FFQ asking about frequency of consumption of a list of foods.
- Adjustment
- Adjustment is often performed in a statsical analysis on an observed effect, commonly because of differing distributions of age, sex, education or other known factors in the populations being compared that may be confounders.
- Aetiology
- The cause of disease.
- ANOVA - Analysis of Variance
- Calculates if there is a statistical difference in the mean values between independent (unrelated groups).
- A posteriori
- Knowledge and reasoning obtained from information/data that has been previously gathered in the study e.g. dietary pattern analyses where researchers already have relevant exposure data to hand.
- Applicability of Tool
- How relevant the tool is for your population; and how easy to use and administer in practice.
- A priori
- Knowledge and reasoning used before the study is conducted e.g. a prospective/predictive power analysis is used to determine sample size or a research question is predetermined/set before the study is conducted.
- Association
- Statistical dependence between two or more events, characteristics, or other variables. The presence of an association does not necessarily imply a causal relationship.
- Attenuation
- The reduction in association or effect size that can occur due to measurement error. For example, when comparing two different methods the measured agreement might be weaker than the true agreement as a result of measurement errors in one or both methods.
- Attrition
- How many people drop out of the study over time during the study period, compared to the number at the start. Attrition rate is usually calculated on study completion. Often expressed as a percentage. It can introduce bias if the characteristics of people lost to follow-up differ between groups.

### B

- Between-Subjects Variation
- Differences between individuals for a given variable of interest e.g. habitual intake. Also known as inter-individual variation.
- Bias
- The level of deviation of results from the true value/effect you are trying to measure; deviation is systematic (as opposed to random) and can be in either a positive or negative direction. Bias/systematic error therefore can lead to an incorrect measurement of the effect or association. Many factors which effect the collection, analysis, interpretation, publication or review of the data can lead to systematic error and biased results of the study. Two main types are selection bias (how subjects were selected) and information bias (how information was obtained). There are many strategies used to reduce the effect of bias such as: multiple control groups, standardised measurement techniques, collaboration of multiple information sources.
- Bioavailability
- The extent to which a nutrient can be taken up by the body and that is available for use and storage. The bioavailability of a nutrient may be affected by interaction with other nutrients.
- BIOBANK
- A collection of human tissues/samples (Allows medical information about donors to be collected to be used for a research purpose).
- Biomarker (Biological Marker)
- Genes or characteristics which are typically components of body fluids (blood or urine) or tissues. The variable can be used to estimate true dietary intake and the extent of exposure of particular nutrients, by detecting the amount of biomarker available in the sample. Dietary biomarkers are of 3 types: recovery, predictive and concentration. Recovery biomarkers (e.g. urinary nitrogen, potassium & sodium) provide an estimate of absolute nutrient intake levels over a fixed period of time. Predictive biomarkers (e.g. fructose, glucose) can be used to assess the validity of a dietary assessment method but have lower overall recovery. Concentration biomarkers (e.g. plasma vitamin C) are strongly correlated with true intakes but cannot be translated into absolute levels of intake. There are only a limited number of nutritional biomarkers available to measure nutrient intake. Biomarker values are gold standard methods to compare intake recorded by a dietary assessment method.
- Bioshare
- Biobank Standardisation and Harmonisation for Research Excellence in the EU.
- Bland-Altman Plot
- A method of data plotting used to analyse the limits of agreement between two different dietary assessment tools. The difference between the two measures is plotted on the y-axis against the average of the two measures on the x-axis. Please find further information in the help section: Statistical tests used in validation studies
- Burden
- The additional effort and time for both the researcher and the participant to complete the dietary assessment.

### C

- Calibration
- A mathematical measure to calculate the measurement error between two methods, the values from one method are quantitatively related to values from a reference/standard method. The process of re-scaling data obtained from a more biased, less accurate instrument based on information obtained from a less biased, more accurate instrument.
- Case
- A single participant in a case control study. Participants who develop the outcome of interest in cohort studies may also be referred to as cases.
- Case-Control Study
- An observational study where participants are grouped by disease (outcome) and their different exposures and frequency of the exposures are measured. A retrospective study; the outcome of interest is measured in the case group (participants with condition/disease), the controls are identified (no disease or condition but share similar characteristic and are at risk of developing the disease) and the exposures of different factors to both groups are measured to identify whether a cause of the outcome can be established. The risk of developing the condition/disease based on the level of exposure is then calculated. This study design is particularly prone to recall bias.
- Cause
- An exposure (e.g. diet) that is influential in the occurrence of the outcome (e.g. disease). Observational studies (cohort studies, case-control studies) assess associations between exposure and outcome, but associations do not necessarily imply a causal relationship. Experimental studies may get closer to causal relationships. Bradford Hill proposed 9 factors to be necessary to provide adequate evidence of a causal relationship between an exposure and outcome.
- Cluster Analysis
- A group of participants are tested and measured and group level data is compared rather than individuals. For example, a study may work with clusters such as school classes or general practise populations.
- Co-efficient of Variation
- A measure of reproducibility. The ratio of the standard deviation to the mean. This is meaningful only if the variable is measured on a ratio scale.
- Cohen's Kappa Coefficient
- A statistical measure of the level of agreement between two different measures of the same concept used to capture test or retest or inter rater reliability. It used for nominal variables and is analogous to the correlation coefficient for used for continuous variables (and has the same range of values (-1 to +1)). Kappa tests the inter-rater reliability and it is the adjusted agreement between the observed proportional agreement, which takes into account the amount of agreement that would expected by chance.
Please find further information in the help section 'Statistical tests used in validation studies'.

- Cohort Group
- An identifiable population group who are defined by a particular characteristic/experience such as geographical location, year of birth or occupation. They are followed over time and their outcomes are measured.
- Cohort Study
- A prospective study where current exposure of a factor/factors to the individuals within the group are measured. The participants are followed into the future and the frequency, distribution and occurrence of outcomes are measured. (A Follow Up or Longitudinal Study). Those who are exposed and unexposed to the factor/factors are compared to identify those who develop the outcome/s of interest (disease). Cohort studies are able to measure numerous outcomes but are limited to the extent of exposure they can measure. Can also be retrospective where records of past exposure can be reviewed to determine their effect on current health.
- Confidence Interval
- The range of a measured value, which contains the true population mean which is calculated by using ±2 standard deviations of the mean value. It is conventional to create confidence intervals at the 95% level – so this means that 95% of the time properly constructed confidence intervals contain the true value of the variable of interest. This corresponds to hypothesis testing with p-values, with a conventional cut-off for p of less than 0.05.
- Confounders
- A type of bias. Confounders alter the relationship/association we are trying to measure/detect between the exposure and outcome. They are variables which are associated with both the exposure and outcomes of the study. They are associations which can lead to over/under estimation of the true measures of effect. The effect of confounders can be minimised at both design stage (randomisation, restriction, matching) and at analysis stage (stratification or adjustment for known measured confounders).
- Construct Validity
- The ability of the measurement tool to actually measure the construct it claims to be measuring or the concept being studied (e.g. association between a food environment tool and dietary intake).
- Content Validity
- How the content of the tool (items tasks, format and wording of the tool) and, the processes required of the participants appear to relate to the content being measured. The tools content is evaluated by experts to ensure that the tool provides sufficient, clear and relevant results.
- Control
- A single participant with similar characteristics/or matched on certain criteria to the case in a case-control study.
- Control Groups
- A comparison group which will have similar characteristics/certain criteria to the case group but have not been exposed to the intervention or factor of interest.
- Convergent/Concurrent Validity
- The extent of agreement between the tools compared, how the tool relates to other similar tools measuring the same or similar constructs. A high correlation between the test and the reference/standard tool supports the validity of the test.
- Correlation Coefficient
- Used to assess the linear association between the test method and the reference method. However, the correlation coefficient does not imply causality. It is denoted by the symbol 'r'. The correlation coefficient is always between -1 and 1, thus -1 < r < 1.
Please find further information in the help section 'Statistical tests used in validation studies'.

- Covariate
- A variable that is possibly predictive of the outcome under study. The covariate may be either a direct interest to the study or a confounding variable or an effect modifier.
- Criterion Validity
- The extent to which the measurement correlates with an external criterion of the phenomenon under study, e.g. reference method. There are two aspects of criterion validity which are (i) Concurrent validity, where the measurement and the criterion refer to the same point in time, and (ii) Predictive validity, where the measurement’s validity is expressed in terms of its ability to predict the criterion.
- Cross-sectional Study
- An observational study which measures the relationship between both the exposure of interest (diet) and outcome (disease) (neither of which are known or specified at the start of the study), at one particular point in time. Individuals must be representative of the whole population. Studies provide a snapshot of a population therefore cause and effect cannot be distinguished.

### D

- DANTE
- Diet And Nutrition Tool for Evaluation (DANTE). DANTE incorporates the McCance and Widdowson composition of food tables. "This database can be used to calculate the total and daily costs estimated from coded dietary intake data by allocating a price to each food or beverage consumed".
- Dependent Variable
- Another name for the outcome variable. The variable (e.g. disease) that is hypothesized to be affected by (dependent on) the exposure variable (e.g. diet).
- Dietary Assessment Tool (DAT)
- These tools are used to collect food intake and include Food Diaries, 24 Hour recalls and Food Frequency Questionnaires
- Dietary Habits
- The habitual intake, namely food patterns, meals and nutrients which are regularly consumed.
- Diet Diversity Score
- Number of food groups or number of foods consumed regularly. Diet Diversity reflects the adequacy of essential nutrients.
- Diet History
- Combination of short-term and long-term methods, usually 24hr recall, FFQ and food diary; more often used in a clinical setting by experienced dieticians to provide in-depth assessment at individual level.
- Diet Record
- Also known as food diary.
- Direct Observation
- Participants intakes are monitored and recorded by trained observers/researchers.
- Discriminant Validity
- Low correlation with measurements/factors that, in principle, they should correlate poorly with.
- Dose-Response
- The relationship between the observed responses or outcomes (e.g. disease) to varying levels/frequencies of the exposure (e.g.diet).
- Doubly Labelled Water (DLW)
- Water with identifiable isotopes of hydrogen and oxygen that can be used as a recovery biomarker for measuring energy expenditure. DLW is assumed to be equal to energy intake in weight-stable individuals.

### E

- Ecological Fallacy
- Inappropriate conclusions are drawn about individuals from measurements made between the exposure of interest (diet) and the outcome measured (disease) at the group level rather than at individual level.
- Ecological Studies
- An observational study which measures population groups rather than an individuals, such as groups from specified locations (cities, or counties) or particular age groups, individuals can not be identified. It used to measure the occurrence of the outcome (disease) in relation to characteristics and exposures to the population.
- Effect Modifiers
- An effect modifier interacts with the exposure of interest which leads to a different outcome.The effect of the primary exposure on an outcome differs depending on the level of a third variable (the modifier).
- Effect Size
- An objective measure of the magnitude of an observed effect e.g. r of Pearson's correlation coefficient.
- Epidemiology
- The study of the distribution and determinants of health states or events in specified populations, and the application of this study to the control of health problems.
- EURRECA EURopean RECommendations Aligned & (Network of Excellence)
- The Network (2007-2012) addressed the problem of variations in micronutrient recommendations between different European countries. Its main objective was to develop methodologies to standardise the process of setting micronutrient recommendations.
- Exposure
- Potential causal characteristic i.e. dietary habits/intakes (Food patterns, meals, foods and nutrients) of the individual, group or population studied.

### F

- Face Validity
- The extent to which a measurement or a measurement tool appears reasonable on superficial inspection.
- False Negative
- Refers to a negative test result in a person who possesses the attribute.
- False Positive
- Refers to a positive test result in a person who does not possess the attribute.
- Food Diary/Record
- Prospective, short-term methods where details of all foods and drinks are recorded by the participant as they are consumed, usually over several days. Amount of food eaten can be either estimated using household measures (estimated food diary) or weighed by the respondent or research assistant in the home (weighed food diary). It can be long-term method if carried out multiple times, i.e. over multiple phases. Also known as food records or diet records. Can be completed online (see emerging technologies section).
- Food Frequency Questionnaire (FFQ)
- Retrospective methods querying frequency over periods of time, questions relate to the frequency with which foods and drinks have been consumed over a long time period (weeks, months, and years). Can be ‘qualitative’ (frequency only), ‘semi-quantitative’ (estimated portion pre-assigned e.g. small, average, large) or ‘fully quantitative’ (portion size queried). Can be long (comprehensive, around 100 items queried or more) or short (also known as ‘screeners’ or a type of brief instrument). Can be interviewer- or self-administered, completed on paper (with potential scanning option) or online (see emerging technologies).
- Food Propensity Questionnaire
- A type of FFQ that does not include portion size information. This FFQ has less respondent burden compared to other FFQ.

### G

- Generalizability
- The degree to which results of a study may apply, be relevant, or be generalized to populations or groups that did not participate in the study.
- Goldberg Equation
- An equation that uses basal metabolic rate (BMR) and physical activity level (PAL) (or estimates of these) to identifying potential under and over-reporters of dietary intake. BMR can be estimated from age, weight and height.
- Gold Standard
- A method, procedure, or measurement that is widely accepted as being the best available. Often used to compare with new methods of unknown effectiveness.

### H

- Habitual Intake
- Usual/long-term dietary intake. There are no methods that measure accurately habitual intake level of single individuals, therefore, differences between measured and true intake levels cannot be evaluated for separate individuals, but only be estimated.
- (data) Harmonisation
- Data is harmonised to achieve or improve comparability of similar measures collected by separate studies. Variables are combined often by simplification of variables to the 'lowest common denominator'.
- Hawthorne Effect
- The effect (usually positive or beneficial) of being under study upon the persons being studied; their knowledge of the study often influences their behaviour.
- Healthy Eating Index
- The Healthy Eating Index (HEI) is a measure of diet quality that assesses conformance to the Dietary Guidelines for Americans.
- Hypothesis
- A prediction of a relationship that is phrased in such a way as to allow it to be tested and confirmed or refuted.

### I

- Independent Variable
- The predictor or explanatory variable, the exposure. An independent variable is not changed, affected or dependent on other variables you are trying to measure but may affect the dependent variable.
- Information Bias
- Occurs where there are either random or systematic differences in the way information is obtained regarding the measurement of exposures or outcomes. Information can be biased if different interviewers are used in the study or the tool does not accurately measure intake. Information bias can lead to participants being misclassified into the wrong exposure group.
- Inter-Individual Variation
- Differences between individuals for a given variable of interest. Also known as between-subject variation.
- Internal Reliabilty
- It refers to the consistency of the measure within itself. The reliability coefficient is used to measure the common characteristics of a test i.e. the extent to which the different questions within a tools are similar.
- Internal Structure Validity
- The extent to which the internal components of a test match the defined measurement of interest (what we aim to measure).
- Inter Quartile Range (IQR)
- The range of the middle 50% of an ordered set of observations, containing the second and third quartiles. The difference between the 25th and 75th percentile values. A measure of dispersion.
- Inter-Rater Reliability
- Extent to which different researchers or observers of a given measure record the same value of the measure for a given case. It is the consistency of results obtained between different observers.
- Interval Data
- The data set is ordered and the size of the differences/values between defined groups are constant.
- Intervention Studies
- A study where an investigator/researcher actively intervenes with participants or a population group. Participants are exposed to an intentional change or treatment and compared to a control group who receive no treatment. Experimental studies such as randomised controlled trials are intervention studies.
- Interviewer Bias
- Systematic error due to interviewers’ subconscious or conscious gathering of selective data or influencing subject response. May occur when researchers ask questions or record data differently between different individuals in the study group, or when a number of different researchers are used to conduct interviews or when some individuals record their data using structured questions and some individuals are interviewed which may lead to more information being revealed.
- Intra Class Correlation Coefficient
- A measure of validation which is used to assess whether similar results are obtained from the dietary assessment tool as from the reference method/tool. It determines the level of association between the two methods, it takes into the account both the degree of correlation and the size of agreement between pairs. Intra class correlation coefficients take into account that associations can occur by chance.
- Intra-Individual Variation
- The level of difference between the mean for each individual. Variations in the values around the individuals true mean when repeated measures are performed with a valid measurement tool, for example variations in an individual's day-to day nutrient intake. Also known as within-subject variation.

### J

### K

- (Cohens) Kappa Coefficient
- A statistical measure of the level of agreement between two different measures of the same concept used to capture test or retest or inter rater reliability. It used for nominal variables and is analogous to the correlation coefficient for used for continuous variables (and has the same range of values (-1 to +1)). Kappa tests the inter-rater reliability and it is the adjusted agreement between the observed proportional agreement, which takes into account the amount of agreement that would expected by chance.

### L

- Limits of Agreement
- A measure of the agreement between two values of repeated measures, often used for validating one dietary assessment tool against another. The difference between the two measures is calculated for each subject, it is then assumed that 95% of the differences between the two measures lie within ±2 standard deviations of the mean difference. This interval is called the limits of agreement (LOA) and is displayed in Bland-Altman plots which plot the difference between the two measures on the y-axis against the average of the two measures on the x-axis. If this interval is considered to be small then the two measures are said to provide similar results. This is the preferred method of measuring agreement of continuous variables.
Please find further information in the help section 'Statistical tests used in validation studies'.

- Longitudinal Study
- A study which occurs over a long period of time, with repeated measures.
- Long-Term Dietary Tool
- A dietary assessment tool which captures information over a long period of time e.g. FFQ.

### M

- Matching
- It is the process of making the distribution of the matching factors similar to the study groups. The selecting of participants to control for confounders, i.e. controls are matched to cases on certain characteristics which could affect the results of the study and the association/relationship you are trying to detect between the exposure and outcome. Matching/controlling for too many variables may mean the study sample may not be representative of the whole population (not externally valid).
- Measurement Bias
- May over or under estimate values and may be caused by biased equipment, analytical bias, social desirability bias (respondents underestimating their true intake), interview bias or recall bias. Refers to a systematic error (bias) in a measurement or a systematic error that occur from inaccurate measurements or classification.
- Measurement Error
- The difference between the measured exposure (measured dietary intake) and the true exposure, the smaller the measurement error the greater the reproducibility of the measurement tool.
- Median
- The middle score of a set of observations. Often reported instead of / or as well as the mean (and standard deviation) for skewed distributions, along with the interquartile range (IQR).
- Missing Data
- Data/values that are missing for individuals who have completed the study where other information has been obtained.

### N

- NIHR
- National Institute of Health Research.
- Nominal Data
- Categorical data which is not ordered e.g. Y/N.
- Non Parametric Tests
- Statistical tests that do not rely on the assumption that the data has a normal distribution.
- Non-Response Bias
- This occurs if there is a systematic difference in characteristics between responders and non-responders.
- Normal Distribution
- A unimodal, symmetrical bell shaped curve where approximately 95% of the values lie within ±2 standard deviations of the mean value.
- Null Hypothesis
- A statement that the predicted effect in the hypothesis is not true e.g. a null hypothesis would propose no variation exists between the measurements from two dietary assessment tools which are compared within a validation study; whereas the hypothesis proposes there is a difference.
- Nutritional Epidemiology
- The study of how diet and nutritional status affect the prevalence and distribution of disease and health conditions.
- Nutrient Density
- Nutrient intake divided by total energy intake.

### O

- Observational Study
- Epidemiological study in which the investigator collects data on exposures and outcomes of interest without attempting to alter the outcome or exposure status of the subjects.
- Odds Ratio (OR)
- Measures the association between the exposure and the outcome (expressed as a binary variable). The odds ratio represents the odds that an outcome will occur given the exposure compared to the odds of the outcome occurring without the given exposure.
- Omissions
- Forgotten items (food and drinks) when recording dietary intake.
- Opal
- Database software for epidemiological studies with biosamples. This database can be used to manage heterogeneous and data from different sources.
- Ordinal Data
- Data that is ordered, there is a difference between values but the level of difference cannot be calculated, i.e. really like, like, no preference, dislike, really dislike.
- Outcome
- The variable (e.g. disease) that is hypothesized to be affected by (dependent on) the exposure variable (e.g. diet). Alternatively, energy intake (kcal or KJ), energy density, nutrient intake, dietary patterns may be outcome variables predicted by variables such as socio-demographic factors. An outcome may be predicted from one or more predictor variables.
- Outliers
- An observation that lies outside the overall pattern of the normal distribution and is distant from the other values. They are so high or low that they can have undue influence on analysis results because they are not representative of the sample.

### P

- Parametric Test
- Parametric tests assume that the data has a normal distribution.
- Pearson Correlation
- A parametric test that investigates the strength of the linear relationship between two quantitative continuous variable. It is used to measure reproducibility of dietary assessment methods e.g. nutrient intakes assessed on the same individual on two different occasions. However the measure of the Limits of Agreement using a Bland-Altman plot may be more appropriate.
- Percentage Agreement
- The percentage of subjects categorised into the same category of exposure when measurements are recorded on separate occasions or by different raters or by different methods; used for determining the reliability and validity of dietary assessment methods. Cohen's Kappa coefficient can be used to calculate the amount of agreement taking account of the amount of agreement that would be expected by chance.
Please find further information in the help section 'Statistical tests used in validation studies'.

- Percentage Variance
- The percentage of the total variance among the variables accounted for by each factor.
- Precision
- The extent of the agreement between repeated measures of the same variable under unchanged conditions.
- Predictive Validity
- The extent to which a tool correctly predicts outcomes e.g. presence/ absence of disease mortality or body composition, which can be measured with more certainty later.
- Prevalence Rate
- Measures the burden of disease in a population/group within a specified time. It includes all individuals who have the disease including long term sufferers and newly diagnosed cases.
- Prospective
- "An observation which occurs in the present i.e. recording individuals current dietary intake. Individuals record their intake at time of consumption, prospective studies are able to record variation in daily diets."
- Prospective Study
- Prospective study follows the outcome development over a long period of time. Prospective studies usually have fewer potential sources of bias and confounding compared to retrospective studies as confounders are considered before the study begins.
- Population
- In statistical terms it refers to a group of individuals or units to which statistical findings can be generalised.
- Power
- The probability of detecting an effect of a particular size, given that the effect is present. It is the probability of rejecting the null hypothesis when the alternative hypothesis is true.
- Power Calculation
- How likely the study is to produce a statistically significant result; the probability that a test correctly rejects the null hypothesis. Power calculation is used to determine what the desired sample size of the study would be in order to detect a statistical difference.
- P-value
- The probability associated with the outcome of interest. It is the probability that the result occurs by chance, if the null hypothesis is true (i.e. that there is no difference or no relationship between variables). The results are considered real and unlikely to be caused by chance when the p-value is <0.05, being statistical significance.

### Q

- Qualitative
- The measurement and analysis of observations that cannot be quantified.
- Quantitative
- The measurement and analysis of observations in a systematic and numerical way.

### R

- Random Error
- Errors that are difficult to predict. Deviations from the true population value due to chance, which can be reduced by increasing the sample size. Includes individual biological variation, sampling error and measurement error e.g. measurement error can be caused by incorrectly measuring the amount of food consumed. To reduce random error standardised measurement techniques, trained personal or quality control procedures need to be used.
- Randomisation
- Assignment of individuals at random to experimental and control groups. Randomisation limits biases, conscious or unconscious, that may influence the allocation of participants to experimental and control groups. Successful randomisation results in groups which are similar in confounding characteristics (confounders).
- Randomised Control Trial (RCT)
- Where all participants are randomly assigned to one of the treatments/interventions to be tested or to the control group. In an intention to treat (ITT) analysis of an RCT, participants who were allocated to the intervention are analysed together regardless of whether they completed that treatment or received it. This ignores non-compliance, protocol deviations, withdrawal and anything that happens after randomisation.
- Random Sample
- A group of subjects selected from a population in a random manner (i.e. each member of the population has an equal chance of being selected).
- Range
- A measure of dispersion from the lowest and highest values.
- Reactivity
- The measurement of change in the behaviour due to participants awareness that they are now or will be measured resulting in misreporting (A type of bias).
- Reactivity Bias
- A change in dietary behaviour in some participants due to their awareness that it is being or will be measured.
- Recall Bias
- A type of systematic error caused by lack of accuracy or incompleteness of information which participants remember and record from the past. Participants can not accurately remember information in regards to what they ate when "recalling" their diet from the previous day or reference period and therefore important information is often missed. Systematic error can be reduced by trying to gather independent references i.e. medical records or receipts.
- Reference Method/Tool/Instrument
- A method that is used to calibrate or validate the main dietary assessment tool to be used in a study. The reference method is assumed to provide estimates that are closer to truth than the main tool and is usually a weighed diary or biomarker. May also be called the standard method.
- Regression Analysis
- May be used to explore the relationship between the dependent (outcome) variable and independent (predictor) variables. Multiple regression analyses use more than one predictor variable.
- Relative Risk
- The ratio of the rate of appearance of the outcome (disease) measured in the population group which have been exposed to different levels of the exposure of interest (e.g. low intake group, high intake group). The incidence of outcome in the exposed divided by the incidence of outcome in the unexposed. It is usually a synonym for risk ratio, but this term is also used to refer to the rate ratio and to the odds ratio (OR).
- Relative Validity
- Relative validity is determined by comparing a dietary assessment tool with an alternative dietary assessment method with its own limitations i.e. that is not the gold standard method.
- Reliability
- The degree of stability of a measurement repeated under the same conditions.
- Repeatability
- Used to help measure reproducibility of the study: the variation in results of repeated measures (at least 2 measures) made on a sample of subjects in identical conditions. Where the same measurement instrument, method and same observer are used and the test is repeated over a short period of time. This identifies errors due to measures of processing itself.
- Reproducibility
- The level of variation on the results obtained from subjects under changing conditions, such as different tools, different raters, or different time periods.
- Residual Confounding
- Occurs when a confounder has not been adequately adjusted for in analyses.
- Response Bias
- This occurs if there is a systematic difference in the way that respondents answered questions, so that their answers did not accurately represent their experiences.
- Response Rate
- The proportion of people who respond positively to the invitation to the study.
- Responsiveness
- The ability of an instrument to identify change accurately. A measure of change to exposure in individuals over time, usually following intervention or treatments.
- Restriction
- Excluding certain participants because of certain characteristics or health status. i.e. exclude smokers.
- Retrospective Recall
- A method in which participants recall information from past events i.e. past dietary exposures. These can be used in retrospective and prospective studies.
- Retrospective Study
- A research design in which participants are classified as some having the outcome (disease cases) and others lacking of it (controls). Case-control studies are retrospective studies. Studies where data which is collected from the past, i.e. past exposures.
- Reverse Causation (Reverse causality)
- This refers either to a direction of cause-and-effect contrary to common assumptions, or to a two-way causal relationship i.e. where the outcome (disease) affects level and frequency of exposure e.g development of disease may change the concentrations of a nutrient biomarker.
- Risk
- The probability that an event will occur within a period of time.
- Risk Factor
- Any lifestyle factor or exposure of interest which can increase the risk of developing a non-communicable disease. Risk factors are variables that are associated with the development of an outcome.

### S

- Sample Group
- Subset or representative group of the total population group.
- Sample Size
- The number of individuals who make up the sample group, the larger the sample size the more likely the results will be representative of the population they represent.
- Season Effect
- Variation in an individuals' diet due to differing seasonal consumption patterns.
- Selection Bias
- Selection bias refers to the systematic differences between the characteristics of the study population and the target population, or systematic differences between the characteristics of the intervention group and the control group. Selection of participants prior to the study commencing can introduce bias into the results if the sample group which are measured are different to the population which they should represent. The measurements recorded between the exposure of interest (diet) and the outcome measured (disease) in the study group are therefore different to the results that may be obtained from participants who are eligible to participate but are not selected.
- Self-Reported Intake
- Dietary intake as recorded by the individual.
- Sensitivity
- Extent to which a diagnostic test or marker correctly identifies those who have a particular condition or disease (true positives). For a person with a disease or condition an effective sensitivity test or marker should always identify the individuals as having the disease.
- Short-Term Dietary Tool
- Captures intake over a short period of time of less than a week such as a food diary/ dietary record and a 24 Hr Recall.
- Skewed Distribution
- The distribution of results is not symmetrical, more observations/values fall on one side of the mean than another. Can be positively or negatively skewed.
- Social Desirability Bias
- Under or over reporting of dietary intake or lifestyle factors to avoid criticism/judgement by the researcher.
- Spearman Rank Correlation
- Non parametric correlation coefficient measures the strength of a relationship between two variables that are not normally distributed. It compares the ranking of the two variables and therefore is suitable for both continuous and discrete variable, including ordinal variables (Non parametric equivalent to Pearson’s correlation coefficient). However, the measure of the Limits of Agreement using Bland-Altman plot may be more appropriate.
- Specificity
- Extent to which a diagnostic test or marker correctly identifies those who do not have a particular condition or disease (true negatives). For those without a condition a specific test should always be negative and identify the individuals as disease free.
- Standard Deviation
- This is a calculation which determines the level of variation of individual values which surrounds the mean.
- Standard Error
- This is a measure of the uncertainty in a sample statistic; the standard deviation of the sampling distribution is the standard error. It is used to calculate confidence intervals.
- Standardization
- The process of converting a variable into a standard unit of measurement allowing the comparison of data with different units of measurement. The unit of measurement typically used is standard deviation units (z-scores).
- Stratification
- Dividing/splitting the study group into smaller groups of individuals which share a characteristic i.e. males and females. Compare the different strata to the sample population to identify if the results detected are the same.
- STROBE-nut
- STrengthening the Reporting of OBservational studies in Epidemiology. Guidelines that aids researchers on reporting findings on nutritional epidemiology studies.
- Systematic Error or Bias
- Causes result to vary from the true value (includes selection bias and measurement bias), the error occurs consistently or in a common direction (error is not random) either positively or negatively therefore alters the mean or median. Systematic error is usually caused by problems with measurement. Systematic error cannot be removed by subsequent statistical analysis therefore needs to removed/minimised during study design.

### T

- Test Method
- The dietary assessment tool which is being validated against a reference/standard method.
- Test-Retest Reliability
- Measures reliability of measurements obtained by administering the same test twice, using the same tool under the same conditions and within a short period of time. The time taken between the two tests should be short enough to avoid changes in behaviour.
- Triad Method
- A triangular comparison, using questionnaire, reference method and biomaker, which assumes independent random error. Measurement of the level of agreement between the results obtained from the test method and the true value relies on correlation between the three measures.
- True Positive
- Refers to a positive test result in a person who does possess the attribute.
- True Negative
- Refers to a negative test result in a person who does not possess the attribute.
- T-test
- Measures the agreement between the means of two sets of results. The independent t-test determines whether two means from independent samples differ significantly, whereas the paired t-test determines whether the means of two related samples, matched samples or repeated measures differ significantly. Usually, the p-value of the two-tailed t-test is reported for a non-directional hypothesis i.e. when no direction of a relationship was assumed.
- 24 Hr Recall
- Retrospective, short-term method where details of foods and drinks consumed over previous 24 hours recalled. Can be administered by an interviewer (face to face or by telephone) following a standardised protocol or can be completed online. Can be administered as a single recall (for group-level assessment) or on multiple days (multiple recall) (required to capture individual variation). It can be used as a long-term method if carried out over multiple phases.
- Type 1 Error
- The error of detecting a relationship/association between exposures and outcomes (rejecting a true null hypothesis of no difference) when there is no real relationship/association between the exposure and outcomes.
- Type 2 Error
- The error of detecting no relationship/association between exposures and outcomes (accepting a true null hypothesis of no difference) when there real is a relationship/association between the exposure and outcomes.

### U

- Under-reporting
- A type of misreporting in which individuals report less than their true intake, e.g. in an FFQ, foods perceived as "bad" are under-reported.
- Usability Study
- A study which evaluates the user experience of a tool such as ease of use.
- Usual Dietary Intake
- Long-term average intake (also called habitual intake).
- Usual Nutrient Intake
- Long-term average nutrient intake. Can be estimated from a dietary assessment tool that measures consumption over a long time period (weeks, months, and years) e.g. an FFQ.

### V

- Validation
- The process of establishing that a method, self-reported instrument or data are correctly measured. Validation or “calibration” evaluates whether a measuring instrument really measure what it suppose to measure.
- Validity
- The degree to which inferences drawn from a study are valid. Validity of a tool can be determined by using a reference method to which the test tool can be compared against. This is known as relative validity as there is little possibility that the reference method used has captured the absolute measure of exposure. Validity has many different dimensions including face, content, construct, criterion and concurrent validity.
- Validity (Content)
- How the content of the tool (items tasks, format and wording of the tool) and, the processes required of the participants appear to relate to the content being measured. The tools content is evaluated by experts to ensure that the tool provides sufficient, clear and relevant results.
- Validity (Convergent/Concurrent)
- The extent of agreement between the tools compared, how the tool relates to other similar tools measuring the same or similar constructs. A high correlation between the test and the reference/standard tool supports the validity of the test.
- Validity (Criterion)
- The extent to which the measurement correlates with an external criterion of the phenomenon under study, e.g. reference method. There are two aspects of criterion validity which are (i) Concurrent validity, where the measurement and the criterion refer to the same point in time, and (ii) Predictive validity, where the measurement’s validity is expressed in terms of its ability to predict the criterion.
- Validity (External)
- The degree to which results of a study may be generalized to another population. How much the information obtained relates to people outside the study sample. How well can the results be generalized and applied elsewhere to the external population, results may not be generalizable if they are only representative of the sorts of the people that take part. To be externally valid the study needs to be internally valid.
- Validity (Face)
- The extent to which a measurement or a measurement method appears reasonable on superficial inspection.
- Validity (Group Level)
- At group level (i.e the sample population) the Dietary Assessment Tools (DAT) is considered to be valid if it provides an unbiased estimate of the mean group intake.
- Validity (Individual Level)
- At individual level the Dietary Assessment Tools (DAT) is considered to be valid if it is able to rank the respondents in the same order or same category (tertiles, quartiles) as the reference method.
- Validity (Internal)
- The extent to which the results of a study are not affected by bias or confounding. Good internal validity is essential for external validity to occur.
- Validity (Predictive)
- The extent to which a tool correctly predicts outcomes e.g. presence/ absence of disease mortality or body composition, which can be measured with more certainty later.
- Variable
- A factor which can be measured or recorded i.e. characterisitcs of the study population.
- Variance
- It is a measure of the dispersion shown by a set of observations, defined by the sum of the squares of deviations from the mean, divided by the number of degrees of freedom in the set of observations. Variance can also refer to the variation of dietary intake between two dietary assessment tools.
- Variance (Between Subjects)
- The level of difference between the mean which is calculated for an average population. Can include variation between individuals' habitual intake, it cannot be measured directly but can be obtained through calculations.
- Variance (Within Subjects)
- The level of difference between the mean for each individual. Variations in the values around the individual's true mean when repeated measures are performed with a valid instrument, for example variations in one individual's day-to day nutrient intake.

### W

- Wilcoxon-Mann-Whitney Test
- Non parametric test that is an equivalent to the parametric independent samples t-test. Wilcoxon Mann Whitney test measures the agreement between the means of two independent samples however, does not assume that the samples are normally distributed whereas the corresponding two sample t-test does. Also called the Wilcoxon's rank-sum test.
- Wilcoxon-Signed Rank Test
- Non parametric test that is the equivalent to the parametric paired t-test. Wilcoxon-signed rank test measures the agreement between the means of two related samples, matched samples or repeated measures however, it does not assume that the samples are normally distributed whereas the corresponding paired t-test does.
- Within-Subject Variation
- The level of difference between the mean for each individual. Variations in the values around the individuals true mean when repeated measures are performed with a valid measurement tool, for example variations in an individual's day-to day nutrient intake.

### X

### Y

### Z

- Z-Score
- The value of an observation expressed in standard deviation units. The distribution of all the z-scores of a population has a mean of zero and a standard deviation of 1.