Statistical analysis of structural parameters of garnet and garnet-like str
uctures retrieved from the Inorganic Crystal Structure Database has reveale
d strong correlations between them. The degree of correlation depends on do
ping site and garnet type. Structural parameters are not equally sensitive
to changes imposed by doping. The strongest correlation exists between the
unit-cell parameter and the doping-ion radius. In the case of doping at the
dodecahedral sites, the most sensitive parameter is the shortest dodecahed
ral distance. Principal component analysis (PCA clearly shows that the firs
t component explains most of the variation of structural data. It has a geo
metric basis and correlates with the effective radii of the doping ion. Par
tial correlation helps to identify the strangest relations between pairs of
variables when the influence of other structural variables on them is cont
rolled. It appears that in some cases partial correlation has a different s
ign when compared with the results of the standard correlation technique.