Glossary

* | Terms : Anthropometry | Instruments | Postures | Landmark | Literature

* | Reliability of anthropometric data : Reliability of anthropometric data | Training of measurers | Errors in anthropometry | Data editing

* | Height and Maximum height

- Anthropometry:
- Traditional methods were used to measure each man in specific postures. Definitions of measurements, landmarks, postures, measurement cloths, and instruments were standardized in order to obtain comparable measurements.
- Instruments:
- The following instruments are used in traditional anthropometry. Measurements are read to the nearest mm.
- Anthropometer
- Sliding caliper
- Large sliding caliper
- Spreading caliper
- Tape
- Postures:
- - Standing posture： The subject stands erect with heels together, arms hanging freely, head orientated in the Frankfurt plane.
- - Frankfurt plane： The Frankfurt plane is defined by the left orbitale (the lowest point on the lower edge of the orbit) and the right and left tragions (upper edge of the tragus). In the postures used for anthropometry, this plane is held horizontal.
- - Sitting posture： The subject sits erect on a hard surface with the thighs fully supported. The thighs are parallel. The shoulders are relaxed. The head is orientated in the Frankfurt plane. When the feet are supported, the height of the foot support is adjusted so that the knees and ankles form right angles.
- Landmark:
- Specific points on the body are defined based on anatomical structures. These points are used to define body dimensions. However, the standardization of landmark names and definitions is incomplete.

1. Martin, R. and R. Knussmann, 1988: Anthropologie. Band I. Gustav Fischer, Stuttgart. (in Germany)

2. National Institute of Bioscience and Human-Technology, 1994: Reference Manual of Anthropometry in Ergonomic Designing. Nippon Shuppan Service, Tokyo. (in Japanese)

3. JIS Z 8500:2002 Japanese Industrial Standard, Ergonomics - Basic human body measurements for technological design. Japanese Standards Association, Tokyo.

4. ISO 7250: 1996: Basic human body measurements for technological design.

*Webpage

- - Reliability of anthropometric data
- 1) sampling bias;
- 2) skill of the measurer(s);
- 3) data editing (see "Data editing")
- - Training of measurers
- - Errors in anthropometry
- - Data editing

The reliability of anthropometric data is primarily determined by three factors:

The sampling bias is the degree to which the measured subjects are not representative of the population. If the skill of the measurer is sufficient, the measurements are repeatable and the values obtained are similar to those obtained by a skilled anthropometrist. Data editing is necessary to eliminate erroneous values caused by various mistakes.

Random sampling helps prevent sampling bias, but most surveys do not include random sampling. When the target population is 100% Japanese, the magnitude of the sampling bias can be evaluated by comparing data obtained for mean height and weight between the survey in question and a national survey conducted by the Japanese government.

See "Training of measurers" for a discussion of the skill of the measurer(s).

The sampling bias is the degree to which the measured subjects are not representative of the population. If the skill of the measurer is sufficient, the measurements are repeatable and the values obtained are similar to those obtained by a skilled anthropometrist. Data editing is necessary to eliminate erroneous values caused by various mistakes.

Random sampling helps prevent sampling bias, but most surveys do not include random sampling. When the target population is 100% Japanese, the magnitude of the sampling bias can be evaluated by comparing data obtained for mean height and weight between the survey in question and a national survey conducted by the Japanese government.

See "Training of measurers" for a discussion of the skill of the measurer(s).

Measurement values obtained from the same subjects by 2 different observers may not be identical; the difference is referred to as "inter-observer measurement error". The main reason for inter-observer measurement error is differences in locating landmarks. In a large-scale anthropometric survey, if different observers take measurements in different regions, inter-observer measurement error can appear as geographic differences.

Inter-observer measurement error can be reduced by standardizing measurement techniques, which can be achieved by appropriately training measurers. In some surveys, to avoid effects of inter-observer error, a single trained measuring team takes all measurements.

When a single observer measures subjects twice, the measurements thus obtained may not be the same; the difference is referred to as "intra-observer measurement error". Intra-observer measurement error can only be reduced by improving measuring skills through training and experience. In a large-scale survey, it may be necessary to monitor the skills of measurers.

Inter-observer measurement error can be reduced by standardizing measurement techniques, which can be achieved by appropriately training measurers. In some surveys, to avoid effects of inter-observer error, a single trained measuring team takes all measurements.

When a single observer measures subjects twice, the measurements thus obtained may not be the same; the difference is referred to as "intra-observer measurement error". Intra-observer measurement error can only be reduced by improving measuring skills through training and experience. In a large-scale survey, it may be necessary to monitor the skills of measurers.

Perfectly precise measurement values are unobtainable in anthropometry. Therefore, the reliability of measurements is usually evaluated using the repeatability measures. Usually, the technical error of measurement (TEM) and/or mean absolute difference (MAD) are used to describe the magnitude of intra-observer measurement errors. TEM and MAD are calculated using the following formulas, where di is the difference between the 2 measurements taken on the i-th subject (i=1,2, …, n).

TEM=√{Σ(d_{i}^{2}/2n)}

MAD=Σ | di | /n

Both TEM and MAD represent the magnitude of the difference between the 2 repeated measurements. For 95% of the time, the difference between the repeated measurements is in the range of ±1.96 × TEM. The main reason for intra-observer measurement errors is variation in locating landmarks.

For data on TEM and MAD of body dimensions measured by skilled anthropometrists, see the following: Kouchi, M. et al., 1996. Random errors in anthropometry. Journal of Human Ergology, 25: 155-166.

TEM=√{Σ(d

MAD=Σ | di | /n

Both TEM and MAD represent the magnitude of the difference between the 2 repeated measurements. For 95% of the time, the difference between the repeated measurements is in the range of ±1.96 × TEM. The main reason for intra-observer measurement errors is variation in locating landmarks.

For data on TEM and MAD of body dimensions measured by skilled anthropometrists, see the following: Kouchi, M. et al., 1996. Random errors in anthropometry. Journal of Human Ergology, 25: 155-166.

The most reliable method for finding erroneous values in an anthropometric database is locating outliers in a 2-dimensional scattergram. To do this for a given measurement item (Item 1 in the following example), first find another measurement item that strongly correlates with it (Item 2 in the following example). Draw a scattergram of subjects using Item 1 and Item 2.

Find outliers in the scattergram. To determine whether it is the value of Item 1 or Item 2 that is erroneous, select several measurement items that correlate with both Item 1 and Item 2 (Items 3, 4, …, n). Normalize the measurement values for Items 1 to n of the outlying subjects, using the mean and standard deviation of each item. Compare the normalized values of n items to determine whether Item 1 or Item 2 has the outlying value. The item with the outlying value is the one that is erroneous.

Find outliers in the scattergram. To determine whether it is the value of Item 1 or Item 2 that is erroneous, select several measurement items that correlate with both Item 1 and Item 2 (Items 3, 4, …, n). Normalize the measurement values for Items 1 to n of the outlying subjects, using the mean and standard deviation of each item. Compare the normalized values of n items to determine whether Item 1 or Item 2 has the outlying value. The item with the outlying value is the one that is erroneous.