pairwise: Pairwise comparisons and ranks of scenarios

In carlopacioni/vortexR: Post Vortex Simulation Analysis

Pairwise comparisons and ranks of scenarios

Description

pairwise conducts pairwise comparisons against a baseline scenario using sensitivity coefficients and strictly standardised mean difference. It also ranks scenarios (and/or parameters when relevant) using these statistics. When yrs='max' (default), VortexR automatically sets yrs to the last year of the simulation.

Usage

pairwise(
  data,
  project,
  scenario,
  params = c("PExtinct", "Nextant", "Het", "Nalleles"),
  yrs = "max",
  ST = FALSE,
  type = NA,
  group.mean = FALSE,
  SVs = NA,
  save2disk = TRUE,
  dir_out = "DataAnalysis/Pairwise"
)

Arguments

data	A data.frame generated by collate_dat
project	The Vortex project name
scenario	The ST Vortex scenario name or the scenario that should be used as baseline if simulations were not conducted with the ST module
params	A character vector with the parameters to be compared, default: c('PExtinct', 'Nextant', 'Het', 'Nalleles')
yrs	The year(s) to be analysed, default: 'max'
ST	Whether files are from sensitivity analysis (TRUE), or not (FALSE, default)
type	Type of ST. Possible options are: 'Sampled', 'Latin Hypercube Sampling', 'Factorial' or 'Single-Factor'
group.mean	Whether calculate the mean of the statistics (SSMD and Sensitivity Coefficient) by group. See details
SVs	A character vector with the parameters to be used to group scenarios, default: NA
save2disk	Whether to save the output to disk, default: TRUE
dir_out	The local path to store the output. Default: DataAnalysis/Pairwise

Details

Pairwise comparisons against a baseline scenario are conducted using sensitivity coefficients (SC, Drechsler et al. 1998) and strictly standardised mean difference (SSDM, Zhang 2007).

pairwise ranks, for each population, the scenarios (and SVs if relevant, see below) based on the absolute value of the statistics (either SC or SSMD) regardless of the sign. That is, the scenario with the absolute SC or SSMD value most different from zero will have a rank equal to '1'. The actual statistics need to be inspected to evaluate the direction of the change.

The Kendall's coefficient of concordance is calculated to test whether the order of ranked scenarios (or SVs if relevant) is statistically consistent across the chosen points in time and parameters (or SVs). For example, if 100 years were simulated, yrs=c(50, 100) and params=c('Nall', 'Het'), the consistency of ranking will be tested across the four raters (i.e. Nall at year 50, and at year 100, Het at year 50 and at year 100). Kendall's test operates a listwise deletion of missing data. However, when data in a whole column (i.e. ranks for a parameter) are missing, the column is removed before the statistic is calculated (See vignette for more information).

It is possible to evaluate the mean effect of a range of values for certain parameters on their outcome variables of interest (i.e. ranking the parameters, rather than scenarios). This is automatically done when the analysis is conducted on with ST=TRUE,type='Single-Factor' and there is more than one SV passed with the argument SVs. Alternatively, it is achievable with a combined use of group.mean=TRUE,SVs. The first argument result in the calculations, following Conroy and Brook (2003), of the mean SC and SSMD for each group of scenarios that have different parameter values. SVs provides the names of the parameters to be considered. Parameters are then ranked accordingly (See vignette for more information).

The parameter values passed with SVs are evaluated at year=0. This is done because these parameters may take value 'zero' if the relevant populations goes extinct. There are cases where Vortex may not evaluate these parameters even at year 0. This may happen, for example, when a population is empty at initialization (i.e. the initial population size is zero), or when K is set to zero at the beginning of the simulation. The user has to make sure that the values for the parameters passed in are correct.

Note that it only makes sense to rank parameters in a ST run when the Single-Factor option is used in Vortex. This is because with Single-Factor, the parameters are modified one at the time (See vignette for more information).

Value

A list of six elements:

• A data.frame with SC values for all scenarios

• A data.frame with SSMD values

• A data.frame with p-values for SSMD values

• A data.frame with the scenario ranks based on SC and one based on SSMD

• The output of the Kendall's test

If group_mean=TRUE there will be six additional elements:

• A data.frame with the mean SC values for each parameter

• A data.frame with the mean SSMD values

• A data.frame with p-values calculated for the mean SSMD values

• A data.frame with the parameter ranks based on the mean SC and one based on the mean SSMD

• The output of the Kendall's test performed on the ranking of the parameters

References

Conroy, S. D. S., and B. W. Brook. 2003. Demographic sensitivity and persistence of the threatened white- and orange-bellied frogs of Western Australia. Population Ecology 45:105-114.

Drechsler, M., M. A. Burgman, and P. W. Menkhorst. 1998. Uncertainty in population dynamics and its consequences for the management of the orange-bellied parrot Neophema chrysogaster. Biological Conservation 84:269-281.

Zhang, X. D. 2007. A pair of new statistical parameters for quality control in RNA interference high-throughput screening assays. Genomics 89:552-561.

Examples

# Using Pacioni et al. example data. See ?pac.clas for more details.
data(pac.clas)
pairw<-pairwise(data=pac.clas, project='Pacioni_et_al', scenario='ST_Classic',
               params=c('Nall', 'Het'), yrs=c(60,120), ST=TRUE,
               type='Single-Factor',
               SVs=c('SV1', 'SV2', 'SV3', 'SV4', 'SV5', 'SV6', 'SV7'),
               save2disk=FALSE)

carlopacioni/vortexR documentation built on May 6, 2022, 12:07 p.m.