Corrigendum

Authors

Errata

This article corrects:

  1. An information-theoretical approach to phylogeography Volume 18, Issue 20, 4270–4282, Article first published online: 18 September 2009

Bryan C. Carstens, Fax: 225-578-9527; E-mail: carstens@lsu.edu

Carstens et al. (2009) presented a method for evaluating models of historical demography under the isolation-with-migration model implemented in the software package IMa (Hey & Nielsen 2007). Our manuscript suggested that phylogeographic inferences could be made with a demographic model-selection framework using AIC scores of the models and information theoretic statistics (Burnham & Anderson 1998). Regretfully, we made a mistake in the calculation of the −ln L values reported in Table 3 (Carstens et al. 2009). This mistake resulted from a misinterpretation of the IMa output; it is not possible to calculate the likelihood of the reduced models using IMa. However, we are still able to compare the relative AIC differences among models using the −log(P) values, which are the maximized posterior density functions of each of the models considered by IMa given the data. Their usage in the calculation of information theory metrics such as the AIC differences (Δi) follows from the justification provided by Hey & Nielsen (2007), who used these values to conduct likelihood ratio tests. While the absolute AIC values (reported in Table 4) are affected by the erroneous calculation of the likelihoods of the reduced models, the relative differences among AIC values do not change when the −log(P) values are used to calculate AIC scores. Consequently, neither the biological inferences regarding the evolution of Pseudofagus idahoensis, or the general approach to phylogeography suggested by Carstens et al. (2009) are compromised.

Table 4.   Information theoretic statistics for each of the IMa models. Shown are models considered by IMa, the number of parameters for each model, its AIC score, AIC differences (Δi), model probabilities (wj) and evidence ratio (Emin,j). All values were calculated following Burnham & Anderson (1998)
ModelkAICΔiwiEmin,i
AAC0029.236200.4714588461
AACDD39.71540.47920.2919643821.61478206
ABC00311.09041.85420.0738201566.386586937
ABC0D411.49442.25820.0492855939.565855122
AACDE411.54722.3110.04675082110.08450412
ABCDD411.68982.45350.0405417311.62897699
ABCD0412.54463.30840.01724344127.34134431
FULL513.494.25380.00669949370.37231972
ABB00115.9996.76270.000545055864.9744769
ABA00216.17746.94110.0004559971033.906887
ABBDD216.22286.98650.0004357581081.928093
AAA00316.64127.4050.0002867431644.184836
ABADD316.9087.67180.0002195952146.942466
AAADD217.47488.23850.0001245973783.860259
ABBDE317.7478.51089.49E-054968.136008
ABADE318.22288.98655.90E-057994.427376
AAADE419.666610.43021.39E-0533867.12301

Ancillary