Synthesized anthropometric data for the German working-age population

The anthropometric datasets presented here are virtual datasets. The unweighted virtual dataset was generated using a synthesis and subsequent validation algorithm (Ackermann et al., 2023). The underlying original dataset used in the algorithm was collected within a regional epidemiological public health study in northeastern Germany (SHIP, see Völzke et al., 2022). Important details regarding the collection of the anthropometric dataset within SHIP (e.g. sampling strategy, measurement methodology & quality assurance process) are discussed extensively in the study by Bonin et al. (2022).
To approximate nationally representative values for the German working-age population, the virtual dataset was weighted with reference data from the first survey wave of the Study on health of adults in Germany (DEGS1, see Scheidt-Nave et al., 2012). Two different algorithms were used for the weighting procedure: (1) iterative proportional fitting (IPF), which is described in more detail in the publication by Bonin et al. (2022), and (2) a nearest neighbor approach (1NN), which is presented in the study by Kumar and Parkinson (2018). Weighting coefficients were calculated for both algorithms and it is left to the practitioner which coefficients are used in practice. Therefore, the weighted virtual dataset has two additional columns containing the calculated weighting coefficients with IPF ("WeightCoef_IPF") or 1NN ("WeightCoef_1NN"). Unfortunately, due to the sparse data basis at the distribution edges of SHIP compared to DEGS1, values underneath the 5th and above the 95th percentile should be considered with caution.
In addition, the following characteristics describe the weighted and unweighted virtual datasets: According to ISO 15535, values for "BMI" are in [kg/m2], values for "Body mass" are in [kg], and values for all other measures are in [mm]. Anthropometric measures correspond to measures defined in ISO 7250-1. Offset values were calculated for seven anthropometric measures because there were systematic differences in the measurement methodology between SHIP and ISO 7250-1 regarding the definition of two bony landmarks: the acromion and the olecranon. Since these seven measures rely on one of these bony landmarks, and it was not possible to modify the SHIP methodology regarding landmark definitions, offsets had to be calculated to obtain ISO-compliant values. In the presented datasets, two columns exist for these seven measures. One column contains the measured values with the landmarking definitions from SHIP, and the other column (marked with the suffix "_offs") contains the calculated ISO-compliant values (for more information concerning the offset values see Bonin et al., 2022). The sample size is N = 5000 for the male and female subsets. The original SHIP dataset has a sample size of N = 1152 (women) and N = 1161 (men). Due to this discrepancy between the original SHIP dataset and the virtual datasets, users may get a false sense of comfort when using the virtual data, which should be mentioned at this point. In order to get the best possible representation of the original dataset, a virtual sample size of N = 5000 is advantageous and has been confirmed in pre-tests with varying sample sizes, but it must be kept in mind that the statistical properties of the virtual data are based on an original dataset with a much smaller sample size.