Analysis of multilevel short-term balance studies
Linearity and significance of the nitrogen balance response to
increasing levels of nitrogen intake should be verified for each individual's
response. For subjects not meeting at least one of these criteria their values
should be eliminated from the pooled data and reported separately, with every
effort made to repeat the study with such subjects to determine whether the
anomalous results were due to experimental error or to some metabolic
difference. A regression line, the mean, and 97.5% confidence intervals for the
zero-balance intercept should be calculated using the pooled
data.