Table 3.

Major assumptions of and questions relating to ordination analyses and their implications in population genetic studies using microsatellites

Method1	Assumption or question	References	Related issues	References
In general	How should results from ordination analyses be interpreted?	–	Identifying biologically important structure among results is an open question	Jombart et al. (2009)
In general	Markers are independent sources of ancestry information	Lawson et al. (2012), Baran et al. (2013)	Correlation of markers due to gametic linkage can distort ordination results and impede interpretation	Patterson et al. (2006), Lawson et al. (2012), Baran et al. (2013)
In general	No missing data	–	It may not be clear how missing values for microsatellite loci should be replaced	This paper
In general	Data are noncompositional, relationships between variables are linear	Jombart et al. (2009)	Each allele at a microsatellite locus is treated as a different marker, which creates groups of compositional data	Patterson et al. (2006), Jombart et al. (2008), Odong et al. (2013)
			Microsatellites not formally incorporated into PCA. Analysis combining %PCA with multiple co-inertia analysis may be required	Laloë et al. (2007)
			Little information available on performing transformations to correct for nonlinearity and/or compositional microsatellite data	This paper
PCA	Depicts allele frequency variance, assumes homogeneous variances among alleles	Jombart et al. (2009)	Allele frequencies need to be scaled, but the choice of scaling method for microsatellites may not be obvious or accessible	Jombart et al. (2009), Odong et al. (2013)
PCA	How do outliers influence results from PCA?	–	Outliers are likely to dominate the results and hamper interpretation of other population structure.	Serneels and Verdonck (2008), Zhang et al. (2009)
PCoA	Depicts distance, assumes distance measure is appropriate for data	Jombart et al. (2009)	Microsatellite mutation model is difficult to infer and is costly to incorrectly specify	This paper, introduction
DAPC	Describes between-population variation only	Jombart et al. (2010)	Depends on assumptions of PCA and chosen clustering method	Jombart et al. (2010)

¹PCA, principal component analysis; PCoA, principal coordinate analysis; DAPC, discriminant analysis of principal components.