Negative bias in estimates of the correlation between children's weight-for-height and height-for-age.

Several analyses of cross-sectional anthropometric survey data have found weak correlations between children's weight-for-height and height-for-age. Random non-sampling error in height measurement can produce a significant negative bias in estimates of their correlation. The size of this bias is quantified here using Monte Carlo simulation techniques. Even when two-thirds of height measurements are within two centimeters of their true value, the median estimate of the correlation is .095 below its true value for the sample. There may be differences in the etiology of protein energy malnutrition, as represented by stunting (low height-for-age) and wasting (low weight-for-height), but inferences based on cross-sectional data alone may be misleading, unless random measurement error is negligeable.