Table 2.

Major assumptions of and questions relating to exploratory clustering methods and their implications in population genetic studies using microsatellites

Method1	Assumption or question	References	Related issues	References
In general	Qualitative (strict) group membership	Xu and Wunsch (2005)	Fuzzy methods allow partial group membership	Xu and Wunsch (2005)
In general	Distance measure is appropriate for data	Felsenstein (2004)	Microsatellite mutation model is difficult to infer and could vary among loci and be costly to incorrectly specify	This paper, introduction
In general	Is clustering method appropriate for sample?	–	Many clustering methods are available but have not been thoroughly tested with microsatellites and/or complex population models	Odong et al. (2011)
In general	Do clustering results accurately depict structure in data or distance matrix?	–	Many methods for cluster validation exist, but are not easily available to population geneticists, have not been evaluated with microsatellites, and are infrequently applied in population genetic studies	Xu and Wunsch (2005), Odong et al. (2011)
UPGMA	Structure is hierarchical	Kalinowski (2009, 2011)	Cannot depict nonhierarchical structure	Kalinowski (2009, 2011)
UPGMA	Constant molecular clock	Felsenstein (2004)	Distorts results when rate of evolution varies among samples	Felsenstein (2004)
NJ	Relaxed molecular clock	Felsenstein (2004)	Allows rate of evolution to vary	Felsenstein (2004)
NJ	–	–	Ties are possible when clustering tips. When individuals are closely related, this can lead to falsely high bootstrap values	Felsenstein (2004)

¹UPGMA, unweighted pair group method with arithmetic mean; NJ, neighbor joining.