Skip to main content
. 2014 Jul 28;27(1):69–83. doi: 10.1002/ajhb.22589

Table 1.

Analytical strategies in growth research and their applications

Strategy Applications
Size Internal Z-scores Standardize a measure for systematic differences between sexes (or systematic differences between any other sub-groups, such as ethnicities)
Standardize a measure for between-child differences in exact age at assessment (using the LMS method)
Transform a skewed measure so that it is normally distributed (using the LMS method)
External Z-scores Compare the mean and distribution of a measure against that in some other sample (typically the reference sample of a growth chart)
Standardize, to some extent, a measure for systematic differences between sexes (when a small sample size prohibits the use of internal Z-scores)
Standardize, to some extent, a measure for between-child differences in exact age at assessment (when a small sample size prohibits the use of internal Z-scores)
Transform a skewed measure so that it approximates a normal distribution (if the growth reference was constructed using LMS or some other technique that adjusts for skewness)
Indices Standardize a measure for between-child differences in total body size (typically taken to be height)
Conditional size measures Standardize a measure for between-child differences in total body size (typically taken to be height)
Growth Conditional regression models Quantify the association of size at one age with an outcome at a second age, conditional on size at the second age (combined with a life course plot to quantify the association of growth between the two ages with the outcome)
Quantify the association of growth between two ages with an outcome at the second age, conditional on size at the first age
Regression with conditional growth measures Quantify the associations of growth during different consecutive age periods with some outcome, conditional on size at the first age
Growth curves Individual growth curves Characterize a child's growth (by fitting a growth curve that summarizes his or her longitudinal data in a few biologically meaningful parameters and/ or derived traits)
Characterize average growth in a sample, after fitting multiple individual growth curves (by producing a mean-constant growth curve)
Characterize between-child and population variation in growth, after fitting multiple individual growth curves (by inspecting the pooled biologically meaningful parameters and/ or derived traits)
Relate growth to some distal outcome, other growth process, or survival process (using a two-step strategy)
Mixed effects growth curves Simultaneously characterize the growth of every child in a sample and the average growth in that sample (by modeling and therefore quantifying within-child and between-child variation)
Quantify systematic differences in growth due to independent variables, such as sex and ethnicity (by adding these variables into the model as fixed effects)
Relate growth to some distal outcome, other growth process, or survival process (using a one or two-step strategy)
Latent growth curves *Same as for mixed effects growth curves*
Patterns of growth Growth mixture modeling *Same as for mixed effects growth curves*
Identify distinct unobserved groups (i.e., latent classes) of individuals who share similar average growth curves
Characterize the determinants of latent class membership and investigate whether or not systematic differences in growth due to independent variables, such as sex and ethnicity, differ across the latent classes
Relate the latent classes to some distal outcome, other growth process, or survival process (using a one or two-step strategy)

LMS, lambda-mu-sigma.