qfasar (version 1.2.1)

find_boot_ss: Find realistic bootstrap sample sizes

Description

In QFASA simulation studies, predator signatures are often simulated by bootstrap resampling prey signature data. find_boot_ss finds bootstrap sample sizes for each prey type that produce simulated predator signatures with realistic levels of variation.

Usage

find_boot_ss(pred_sigs, pred_diets, prey_sigs, prey_loc, n_pred_boot = 1000)

Arguments

pred_sigs

A vector or matrix of predator signatures, intended to be the object pred_sigs returned by a call to the function est_diet.

pred_diets

A numeric matrix of the estimated diet compositions of individual predators, intended to be the object est_ind returned by a call to est_diet.

prey_sigs

A matrix of prey signatures, intended to be the object prey_sigs returned by a call to the function est_diet.

prey_loc

A matrix giving the first and last locations of the signatures of each prey type within prey_sigs, intended to be the object loc returned by a call to the function prep_sig with the prey data frame, or by a call to the function make_prey_part if a partitioned prey library was used for diet estimation.

n_pred_boot

An integer designating the number of predator signatures to bootstrap. See Details. Default value 1000.

Value

A list containing the following elements:

var_diet

A numeric vector of the variance between the estimated diets of all possible pairs of predators, sorted in increasing order.

var_sig

A numeric vector of the variance between the signatures of all possible pairs of predators, sorted consistently with var_diet

.

var_sig_smooth

A loess-smoothed version of var_sig.

mod_sig_var

A numeric vector of the modeled variance between bootstrapped predator signatures at each iteration of the algorithm.

var_target

The target level of variance between bootstrapped predator signatures.

boot_ss

An integer vector of bootstrap sample sizes for each prey type.

err_code

An integer error code (0 if no error is detected).

err_message

A string containing a brief summary of the results.

Details

QFASA simulation studies may require the generation of predator signatures given a specified diet, against which estimates of diet composition can then be compared (e.g., Bromaghin et al. 2015). Given a specified diet, a bootstrap sample of each prey type is drawn and mean prey-type signatures are computed. A predator signature is then generated by multiplying the mean bootstrapped prey signatures by the diet proportions.

Although authors often fail to report the bootstrap sample sizes used for each prey type when describing simulations (e.g., Haynes et al. 2015; Thiemann et al. 2008; Wang et al. 2010), they are subjectively selected (e.g., Iverson et al. 2004; Bromaghin et al. 2015). However, Bromaghin et al. (2016) found that bootstrap sample sizes strongly influence both the bias and the variance of diet composition estimates in simulation studies. Consequently, the selection of bootstrap sample sizes is an important aspect of simulation design.

Bromaghin (2015) presented an objective method of establishing a bootstrap sample size for each prey type that produces simulated predator signatures having a realistic level of between-signature variation. The function find_boot_ss implements the algorithm of Bromaghin (2015). A brief summary of the algorithm, sufficient to understand the objects returned by find_boot_ss, follows below. Please refer to Bromaghin (2015) for additional details.

The concept underlying the algorithm is that the variation in predator signatures can be partitioned into variation due to differences in diet and variation due to prey animals consumed given diet. Consequently, a realistic level of variation between signatures for predators sharing the same diet is approximated from the empirical relationship between a measure of variation between pairs of predator signatures var_sig and a measure of variation between the estimated diets var_diets of the same predator pairs. As the variance in diets approaches zero, the predators are effectively eating the same diet and variation in their signatures therefore approaches the level of variation due to prey selection only. find_boot_ss models the relationship between var_diet and var_sig using a loess smooth, and the modeled signature variance var_sig_smooth for the pair of predators whose value of var_diet is smallest is taken as the target level of variation for predator signatures.

The algorithm is initialized with a sample size of 1 from each prey type. A sample of n_pred_boot predator signatures is generated using those sample sizes and the measure of variance among the signatures is computed. If the variance measure exceeds the target level, the prey type contributing most to the variance measure is identified and its sample size is increased by 1. The algorithm then iterates, increasing the sample size by one at each iteration, until the measure of variation is less than the target level. The level of variation at each iteration and the target level of variation are returned in the objects mod_sig_var and var_terget, respectively.

The argument n_boot_pred should be large enough to return an estimate of the variance measure that itself has low variance, so that the algorithm returns numerically stable results. We suspect that the default value of 1000 errs on the side of caution.

NOTE: Because find_boot_ss is intended to operate on the predator and prey signatures returned by a call to the function est_diet, find_boot_ss can be based on diet estimates obtained in either the predator or prey space, using an original or partitioned prey library. However, it is imperative that the arguments are compatible.

References

Bromaghin, J.F. 2015. Simulating realistic predator signatures in quantitative fatty acid signature analysis. Ecological Informatics 30:68-71.

Bromaghin, J.F., S.M. Budge, and G.W. Thiemann. 2016. Should fatty acid signature proportions sum to 1 for diet estimation? Ecological Research 31:597-606.

Bromaghin, J.F., K.D. Rode, S.M. Budge, and G.W. Thiemann. 2015. Distance measures and optimization spaces in quantitative fatty acid signature analysis. Ecology and Evolution 5:1249-1262.

Haynes, T.B., J.A. Schmutz, J.F. Bromaghin, S.J. Iverson, V.M. Padula, and A.E. Rosenberger. 2015. Diet of yellow-billed loons (Gavia adamsii) in Arctic lakes during the nesting season inferred from fatty acid analysis. Polar Biology 38:1239-1247.

Iverson, S.J., C. Field, W.D. Bowen, and W. Blanchard. 2004. Quantitative fatty acid signature analysis: A new method of estimating predator diets. Ecological Monographs 74:211-235.

Thiemann, G.W., S.J. Iverson, and I. Stirling. 2008. Polar bear diets and Arctic marine food webs: insights from fatty acid analysis. Ecological Monographs 78:591-613.

Wang, S.W., T.E. Hollmen, and S.J. Iverson. 2010. Validating quantitative fatty acid signature analysis to estimate diets of spectacled and Stellers eiders (Somateria fischeri and Polysticta stelleri). Journal of Comparative Physiology B 180:125-139.

Examples

# NOT RUN {
find_boot_ss(pred_sigs = matrix(c(0.05, 0.10, 0.30, 0.55,
                                  0.04, 0.11, 0.29, 0.56,
                                  0.10, 0.05, 0.35, 0.50,
                                  0.12, 0.03, 0.37, 0.48,
                                  0.10, 0.06, 0.35, 0.49,
                                  0.05, 0.15, 0.35, 0.45), ncol = 6),
             pred_diets = matrix(c(0.33, 0.34, 0.33,
                                   0.10, 0.80, 0.10,
                                   0.35, 0.50, 0.15,
                                   0.20, 0.35, 0.45,
                                   0.20, 0.45, 0.35,
                                   0.15, 0.65, 0.20), ncol = 6),
             prey_sigs = matrix(c(0.06, 0.09, 0.31, 0.54,
                                  0.05, 0.09, 0.30, 0.56,
                                  0.03, 0.10, 0.30, 0.57,
                                  0.08, 0.07, 0.30, 0.55,
                                  0.09, 0.05, 0.33, 0.53,
                                  0.09, 0.06, 0.34, 0.51,
                                  0.09, 0.07, 0.34, 0.50,
                                  0.08, 0.11, 0.35, 0.46,
                                  0.06, 0.14, 0.36, 0.44), ncol = 9),
             prey_loc = matrix(c(1, 4, 7, 3, 6, 9), ncol=2),
             n_pred_boot = 500)

find_boot_ss(pred_sigs = matrix(c(0.05, 0.10, 0.30, 0.55,
                                  0.04, 0.11, 0.29, 0.56,
                                  0.10, 0.05, 0.35, 0.50,
                                  0.12, 0.03, 0.37, 0.48,
                                  0.10, 0.06, 0.35, 0.49,
                                  0.05, 0.15, 0.35, 0.45), ncol = 6),
             pred_diets = matrix(c(0.33, 0.34, 0.33,
                                   0.10, 0.80, 0.10,
                                   0.35, 0.50, 0.15,
                                   0.20, 0.35, 0.45,
                                   0.20, 0.45, 0.35,
                                   0.15, 0.65, 0.20), ncol = 6),
             prey_sigs = matrix(c(0.06, 0.09, 0.31, 0.54,
                                  0.05, 0.09, 0.30, 0.56,
                                  0.03, 0.10, 0.30, 0.57,
                                  0.08, 0.07, 0.30, 0.55,
                                  0.09, 0.05, 0.33, 0.53,
                                  0.09, 0.06, 0.34, 0.51,
                                  0.09, 0.07, 0.34, 0.50,
                                  0.08, 0.11, 0.35, 0.46,
                                  0.06, 0.14, 0.36, 0.44), ncol = 9),
             prey_loc = matrix(c(1, 4, 7, 3, 6, 9), ncol=2))

# }