In order to make inference, it is convenient to introduce the additional allocation parameter Z i, with the interpretation that Z i = c indicates that individual i is assigned to mixture component c.
Of course it does not escape the other objection to that inference: it, too, is an inference to an uncheckable conclusion.
With respect to evaluating inference, it is possible we have been too conservative with reporting the performance of our independent training and testing experiments.
Therefore, in order to improve the inference it is necessary to introduce information about the class model in the method.
Tempting as it may be to make this inference, it is crucial not to confuse correlation with causation: perhaps people who were already concerned about climate change were more inclined to make a link to the floods, rather than the other way around.
First, although randomized, placebo-controlled trials are the best approach to establish causal inference, it may not be feasible to conduct one on this topic, which makes the results from a well-designed prospective cohort study with relatively large sample size and long-term follow-up highly valuable.
For extending our work to perform maximum-likelihood inference, it will be desirable to extend the computations to permit the sampling of multiple lineages in each population.
To make a more reliable inference, it is important to use multiple loci, each of which has an independent history (F elsenstein 2006).
In support to our inferences, it is noteworthy that nonorthologous gene displacement has already been documented for recombination functions, as exemplified by the substitution of ruvC by recU or recBCD by addAB (Ayora et al. 2004; Rocha et al. 2005; Cromie 2009).
