We formulated two sets of candidate models.
We have determined two sets of candidate genes located on the chromosomal region of FST (chr5q11.2) and FBN3 (chr19p13.2).
We then generated two sets of candidate habitat suitability models for each species: (i) considering all possible combinations of the selected variables and plausible interactions, and (ii) excluding the three distance variables from the initial dataset of selected variables.
Finally, two sets of candidate genes from genome-wide RNA interference screens are used to test the generality and potential application of the classifier.
The bioinformatics identification of zebra finch prohormone genes was conducted with two sets of candidate prohormone gene lists using the approach described by Southey et al. [ 19, 50].
Comparison of these two sets of candidate genes revealed that almost all (i.e. 99%) of the probesets obtained by our integrated approach were also identified when using a FDR cutoff (Fig. 4).
The simulations were conducted by generating sets of 1,000 mortality time series defined by Model [6] with α = 1.2, one of the confounder specifications A, B, or C, and θ and then by applying BOOT, double BOOT, and BMA using the two sets of candidate models [i.e., with 3 lags of PM2.5 (30 candidate models total) or 1 lag of PM2.5 (10 candidate models total)].
We considered two sets of candidate models that were defined by Model [1] with α taking 10 equally spaced values ranging from α = 0.3 to α = 3, confounders t as defined in specification A, and either three lags of PM2.5 (PM2.5 t,0, PM2.5 t,1, PM2.5 t,2) or one lag of PM2.5 (PM2.5 t,1).
In this example, candidate set 3 has the smallest RMSEs among the three sets of candidate summary statistics.
The two types of RMSEs for the three sets of candidate summary statistics are listed in Table 1.
