What if two PC's are not sufficient
When more than two PC's are required to approximate the two-way table that is to be studied using biplots (i.e., the two-way table after appropriate transformation, centering, and scaling), what should one doWe offer two suggestions.
1) Try to divide the full data into subsets based on the pattern in the biplot, and investigate the subsets in separate biplots. The "Data Stratification" modules provide convenient ways to generate a biplot based on any subset of the full data. Yan and Tinker (2005) have used this strategy in QTL by environment data analysis.
2) Try to examine a rotating 3-D biplot. Such a biplot should be sufficient to reveal the most important patterns in most cases. However, a static 3-D biplot, i.e., a 3-D biplot printed in a 2-D plane, is usually more misleading than helping.