Statistical analyses
Partial Pearson correlations were used to evaluate associations
between maternal and paternal support for PA
and FV and children’s PA and FV behaviour, controlling
for child age, parental age, and parental education. Partial
correlations were calculated for the entire sample
and by gender. To identify groups of families with distinct
configurations of maternal and paternal support for
the target health behaviour, standardized scores on the
maternal and paternal support for PA and FV scales
were entered into a cluster analysis. To determine the
possible number of clusters in the data, we first performed
a hierarchical cluster analysis using Ward’s
method and the squared Euclidian distance as a proximity
measure. The number of clusters was based on the
proximity matrix, fusion coefficients from the agglomeration
schedule, and inspection of the dendrogram and
icicle plots. A plot of the within groups sum of squares
by number of clusters was also used to determine the
appropriate number of clusters [17]. After determining
the number of clusters to test, we then completed a
k-means cluster analysis to: 1) estimate initial cluster
means; 2) determine cluster membership across the entire
sample; and 3) estimate the final solution for the cluster
means. In k-means clustering, k well-spaced observations
from the data are selected at random and used as the initial
estimate of the cluster means. Cases are then assigned
to the nearest cluster based on its distance from the initial
cluster means. After assigning all cases to a cluster, the
cluster means are recomputed on the basis of its member
cases, and the assignment of cases to the nearest cluster is
repeated. The iterative process is repeated until no more
cases move from one cluster to another or the stated
maximum number of iterations is reached. To investigate
the effects of cluster membership on child behaviour,
between-cluster differences on child PA and FV consumption
were tested for significance using oneway ANOVA
with Fisher LSD post-hoc comparisons. Both the cluster
analyses and between-cluster comparisons were completed
using standardised data (sample-based z-scores).
All analyses were performed in IBM SPSS Statistics
(version 21.0).