Australian Bureau of Statistics
4807.0.30.001 - Microdata: National Nutrition Survey, 1995 Quality Declaration
Latest ISSUE Released at 11:30 AM (CANBERRA TIME) 14/06/2013
|Page tools: Print Page Print All RSS Search this Product|
DATA ITEM LIST
SEIFA index of relative social disadvantage
The Index of Relative Social Disadvantage was derived from the 1991 Census. It assigns an index to geographic areas based on socio-economic variables such as economic resources of households, education, occupation, family structure and ethnicity. It is one of five indexes derived by the ABS from the 1991 Census to assist in the analysis of socio-economic characteristics. Details of the indexes are contained in Census: Socio-Economic Indexes for Areas.
Each person on the file was allocated an index score, based on the Collectors District (CD) in which they were enumerated (in most cases their usual residence). The score was grouped by quintile. The quintile relates to the area in which the person was enumerated, not to the socio-economic characteristics of the individual. A high quintile score suggests that the area has fewer families of low income and fewer people with little training and in unskilled occupations, whereas a low score suggests that the area has more families and people of this type.
An indicator of equivalent income for income units has also been placed on each person record on the CURF. The indicator relates to the decile of equivalent income of the income unit to which that person belongs. This has been calculated from the NHS income data collected from all people within a household, and then applied to NNS participants.
The indicator has been derived by summing the individual dollar income of all members of the income unit, and then applying factors from the Henderson Simplified Equivalence Scales. These factors vary according to the composition of the income unit and the labour force status of adult members of the unit. The resulting dollar equivalent income of the unit was then classified by decile. The adoption of the Henderson Simplified Equivalence Scales in this derivation does not necessarily imply ABS endorsement of the scale.
Equivalent income and household income were derived from all income units and households. Approximately 260 NNS participants came from households which did not have complete participation in the NHS (see p. 13). Household and equivalent income has been derived for these households, based on those household members who participated in the NHS. Incomplete households are those with an ABSHID of 20818 or greater.
Foods and beverages consumed in the 24-hour dietary recall were allocated eight-digit codes to uniquely identify each food. The first four digits can be used to categorise foods and beverages into a hierarchical classification system. (This classification has been published in the NNS Users' Guide.) Digits five to seven were used to uniquely identify a food. The last digit of the eight-digit food code indicates whether the food is a modifiable recipe (value of '2') or a single item food/unmodifiable recipe (value of '1'). The file ACBDES.TXT contains a full listing of the food codes and their descriptions (see Appendix 1).
For example, the code for pear juice is 11312801. The first two digits of '11' indicates the fruit juice is a non-alcoholic beverage, the first three digits of '113' indicate that it is a fruit and vegetable juice or drink, and the first four digits of '1131' indicate that it is a single fruit juice. The last digit of '1' indicates that it is a single item food/unmodifiable recipe.
There are two fields which indicate the food code. Most food records will contain an 8-digit code in the field FOODCOD1. However, modified recipes have a 6-digit code in FOODCOD1 and then have the 8-digit code of the base recipe in FOODCOD2. Recipes were generally only modified if the fat or fibre content in the individual's recipe differed significantly from the existing recipe. For most purposes, the 8-digit code of modified recipes is suitable for categorising foods as recipe modifications did not alter the nature of the food/beverage. FOODCOD1 and FOODCOD2 can be combined into a new field NEWCODE using the following criteria:
Plain drinking water
The 24-hour recall collected information on each food and beverage consumed the previous day except for plain drinking water (i.e. tap or uncarbonated bottled water with nothing added). The amount of plain drinking water consumed the previous day was collected as a single amount at the end of the 24-hour recall. Additional information, such as eating occasion or time of consumption, was not collected for plain drinking water. Each person who reported consuming plain drinking water had an extra food/nutrition record added, which contains FOODCOD1 equal to 77777777 and PSZGRAMS equal to the amount drank. No adjustment has been made to the field MOISTUR. Their person record also contains the amount of plain drinking water in the field AMTWATER, and plain drinking water has not been added to TMOISTUR and KMOISTUR.
Combination foods are those where two or more components are combined, usually just prior to consumption, and eaten as a single unit (e.g. tea with sugar and milk). Each member of a combination has a separate food record on the CURF, with these food records having the same value in the field CMBSEQNO (combination sequence number). All foods in the first combination consumed by a person have a CMBSEQNO of 1, foods in the second combination have a CMBSEQNO of 2 and so on.
The field PSZGRAMS on the food record contains the amount of food/beverage consumed. This amount was calculated during data entry from several fields which contained information about the amount eaten.
The Australia New Zealand Food Authority (ANZFA) developed a customised nutrient composition database to enable the 24-hour food records to be converted into nutrient intakes. The database was developed in collaboration with HFS and updated in tandem with the coding of the survey food intakes.
The nutrient composition database was applied to the food intake data and the amount of each nutrient consumed per food per person was calculated (this is at the end of each food record). In addition, the following items were calculated as totals across all foods consumed by each individuals and are on the person record:
Acknowledgment of British nutrient data
Official permission was obtained for the use of up to 4,000 total folate values and general nutrient data for up to 1,000 foods from the UK Ministry of Agriculture, Fisheries and Food and the Royal Society of Chemistry. The following acknowledgment should be provided for any use of the nutrient intake data: "Data from the The Composition of Foods, 5th Edition, and its supplements as produced with the permission of The Royal Society of Chemistry and the Controller of Her Majesty's Stationary Office".
The Food Frequency Questionnaire (FFQ) was offered to people aged 12 years and over on a voluntary basis. The FFQ data has been stored on the person record. There were 9,096 people who returned a usable FFQ and 237 people who returned an unusable FFQ. A person's FFQ was classified as unusable if more than 20 out of the 107 food-lines had a '0' (which means they either didn't respond to the particular food line or gave an unusable response). These unusable records are on the CURF but have not been included in any estimates produced by the ABS.
If you want to analyse people who completed a usable FFQ, then select cases where FFQFLAG=3. A separate set of expansion factors, or weights, was calculated for the FFQ, as participation in the FFQ was voluntary. See pages 26–29 for more information on how to use the weights.
All decimal places have been removed from items on the CURF by multiplying the decimal place out. This needs to be reversed when conducting analysis on these items. The table below details the fields affected and what number the field needs to be divided by.
A large number of variables have codes which indicate either that: the question was not applicable to the respondent (e.g. pregnancy status for males); or the respondent did not answer that particular question. In most cases, these are categorical variables. However, there are some continuous variables on the NNS data files (e.g. height) that used these codes. When calculations such as means, medians or percentiles are done on these variables, you need to exclude the people with not stated or not applicable codes. The variables affected are outlined below.
The item DAYOFWEE indicates which day of week (Sunday–Saturday) the interview was conducted on. The 24-hour recall collected information on intake on the day prior to interview. Therefore, the intake day is one day before the interview day. For example, if the interview day was Monday, then the intake day was Sunday.
The item MINTAKE indicates the month of intake, rather than the month of interview.
Approximately 1,500 people completed the Day 2 NNS. The ABS has used this data to produce a series of factors that can be applied to Day 1 intake to remove the effect of within-person variation on the distribution of one day intakes. The adjusted distribution provides a better indication of the 'usual' distribution of intakes in the population. It is therefore more appropriate for estimating the likelihood of nutrient deficiency or excess in the population when the data are based on only a single day's intake for each person.
A separate set of factors has been produced for each nutrient except alcohol. These factors are in:
Each nutrient has three variables or factors:
mean, calculated from the Day 1 sample;
observed standard deviation, calculated from the Day 1 sample; and
between person standard deviation, calculated from the Day 2 sample.
People wishing to use these factors need to merge them onto the NNS Day 1 person record by age and sex. The factors for each nutrient can be applied to the relevant files using the following formula:
Adjusted value= x + (xi – x) * (sb/sobs),
where x is the group mean for the weighted Day 1 sample,
xi is the individual's day 1 intake,
sb is the between person standard deviation, and
sobs is the group standard deviation for the Day 1 sample.
Methodology used to calculate factors
The adjustment factors were calculated for each nutrient except alcohol. Alcohol was excluded from the analysis because of concerns about estimating within-person variation in alcohol intake from only two days' intakes. It was recognised that an adjustment model based on the entire population was not necessarily appropriate for particular sub-groups. Therefore, adjustment classes based on age and sex were used in the analysis. Although it would have been possible to choose groupings appropriate to each separate nutrient, common groupings was used across most nutrients.
Adjustment factors have been calculated for males and females for the following age groups: 2-3 years; 4-7 years; 8-11 years; 12-15 years; 16-18 years; 19-24 years; 25-44 years; 45-64 years; and 65 years and over. The group mean within each of these age by sex groups was calculated from the total weighted Day 1 sample.
However, collapsed age groups were used to calculate the standard deviation values. This was because the between-person standard deviation was calculated from the replicate sample and therefore some cells had insufficient sample for the finer age groups. The collapsed age groups were: 2-11 years; 12-24 years; 25-44 years; 45-64 years; and 65 years and over. Standard deviation was calculated as the collapsed group standard deviation for the Day 1 sample. The between-person standard deviation was calculated as the collapsed group between-person standard deviation for the Day 2 sample.
Between-person variance (s2b) was calculated from the Day 2 sample, using the SAS ANOVA procedure. Person was the independent variable (each person had a Day 1 and a Day 2 record) and nutrient intake was the dependent variable. Between person variance was estimated as: (s2b) = (MSA - MSE)/n, where MSA and MSE are defined according to the table on the following page.
All calculations for vitamin A expressed as retinol equivalents, preformed vitamin A and provitamin A were done on natural log transformed data, because of the particularly skewed nature of their distributions. A value of 5 was added prior to the log transformation, because there were a small number of zero values and logs are defined for only positive non-zero numbers.
Source: Searle (1971).
Note: SSM, SSA and SST are calculated directly from data collected. SSE is calculated by subtraction.
Structure of the adjustment factors files
The following two column delimited text files have been provided for those people wishing to apply the nutrient adjustment factors to the NNS data file :
These need to be applied differently to the other nutrients. As previously outlined, each nutrient except alcohol has three variables or factors: mean; observed standard deviation; and between person standard deviation. People wishing to use these factors need to merge them onto the NNS Day 1 person record by age and sex.
Note that the factors for the three forms of vitamin A are in the natural log scale. To apply the adjustment factors to the three forms of vitamin A, the following steps should be followed:
1. Add 5 to recorded intakes (to make all intakes non zero numbers).
2. Log transform (base e) the intakes.
3. Apply the factors.
4. Convert the adjusted intakes back into the original scale (ex), and then subtract 5.
Calculations for all other nutrients were done in their original scale.
These documents will be presented in a new window.
This page last updated 13 June 2013