The performance of these criteria on all datasets was estimated with 10-fold cross-validation.
Probability that age-associated methylation changes are related in different datasets was estimated by hypergeometric distribution, and linear correlation was determined by Pearson correlation coefficient.
Propensity scores for each respondent in the filled-in datasets were estimated.
Node support for codon/gene-partioned datasets was estimated by the means of non-parametric bootstrap resampling [ 149].
Usually, microarray datasets are estimated to have more than 5% missing values and up to 90% of genes are affected [ 20, 21].
Divergence ages from both mitochondrial datasets were estimated in BEAST 1.6.1 (Drummond and Rambaut, 2007) with a relaxed molecular clock approach (Drummond et al., 2006).
For ML analysis, the best-fit model of our amino acid sequence datasets was estimated using the Akaike Information Criterion AICC) using ProtTest version 2.0 [ 67].
When hybridization levels for the CON and NEU datasets were estimated using BAGEL, distance phylogenetic analysis recovered the MLSA tree more often, but this advantage was not seen with parsimony analysis.
The expected genome length based on the parental consensus MP1 and MP2 and the Integrated Map datasets was estimated as described in Method 4 [ 53], assuming a random distribution of markers.
Additionally, nodal support of the combined nuclear and plastid datasets were estimated by dividing the number of statistically supported clades a with BS ≥ 85 by the number of possible bipartitions (i.e. the number of terminals minus three).
