The correlation tables are tables of correlation coefficients linking pairs of dimensions. A correlation coefficient is a number between 0 and 1, which conveys the extent to which people tend to be big or small in BOTH dimensions.
A correlation of 0 means that if someone is big in one dimension we still cannot predict at all how big they may be in the other dimension.
Conversely a correlation of 1 means that everyone who is big in one dimension is proportionally big in the other.
In theory, the square of the correlation coefficient explains that proportion of the variation.
In PeopleSize, correlations were derived from the Natick, HUMAG and the UK Department of Health raw data sources, and as quoted by Haslegrave (1981). The most extensive table is from the Natick data, which are of Caucasian US Army personnel, and so contain a source of error in that the people were relatively young and fit. This alters the relationships between bony and fatty dimensions. However it has advantages in that it is a very large dataset with a large number of dimensions, and of very high quality.
The error from the Natick age/shape profile is considered small, compared with errors which otherwise occur through lack of a formal process for estimating Design Percentiles. Most users of anthropometry data treat dimension percentiles as Design Percentiles, through lack of a means of calculating the difference.
Some dimensions in PeopleSize are not in the correlation tables, because the raw data are not available to calculate the coefficients.