calcSplits {GenEst}

Estimate the number of fatalities by up to two splitting covariates

Description

Total mortality can be split into sub-categories, according to various splitting covariates such as species, visibility class, season, site, unit, etc. Given the carcass search data, estimated mortalities, and splitting covariates, calcSplits() gives the "splits" or summaries the estimated mortalities by levels of the splitting covariates. For example, user may specify "season" and "species" as splitting variables to see estimated mortalities by season and species. Input would be arrays of estimated mortalities and arrival intervals when ncarc carcass have been discovered and uncertainty in mortality estimates is captured via simulation with nsim simulation draws.

Usage

calcSplits(
  M,
  split_CO = NULL,
  data_CO = NULL,
  split_SS = NULL,
  data_SS = NULL,
  split_time = NULL,
  ...
)

Arguments

M

estM object, containing numeric array (ncarc x nsim) of estimated mortalities and other pieces

split_CO

Character vector of names of splitting covariates to be found in the data_CO data frame. No more than two split_CO variables are allowed. Use split_CO = NULL if no CO splits are desired.

data_CO

data frame that summarizes the carcass search data and must include columns specified by the split_CO arg. Each row includes search and discovery parameters associated with a single observed carcass. Columns include carcass ID, carcass discovery date, unit, and any number of covariates. data_CO is required if and only if split_CO is non-NULL.

split_SS

Character string giving the name of a splitting covariate in the data_SS list, with data_SS[[split_SS]] describing characteristics of the search intervals (e.g., "season"). Note that length(data_SS[[split_SS]] must equal length(data_SS$days) - 1 because no inference is made about carcass arrivals prior to time t = 0, and the "interval" prior to t = 0 is not taken as a "search interval." If no split_SS split is desired, use split_SS = NULL.

data_SS

Search schedule data

split_time

Numeric vector that defines time intervals for splits. Times must be numeric, strictly increasing, and span the monitoring period [0, max(data_SS$days)]. If no split_time is desired, use split_time = NULL. If split_time is NULL and split_SS is not NULL, data_SS is required.

...

arguments to be passed down

Details

Arrival intervals (Aj) are given as integers, j, that indicate which search interval the given carcass (indexed by row) arrived in the given simulation draw (indexed by column). Arrival interval indices (j) are relative to indexed carcasses' search schedules.

No more than two splitting variables (split_CO, split_SS, and split_time) in total may be used. split_CO variables describe qualitative characteristics of the observed carcasses or where they were found. Some examples include searcher (DHD, JPS, MMH), carcass size (S, M, L), species, age (fresh/dry or immature/mature), unit, visibility class (easy, moderate, difficult), etc.

split_SS variables describe characteristics of the search intervals, such as season (spring, summer, fall, winter) or treatment (pre- or post-minimization). Each search interval is assigned a level of the split_SS variable. For example, for a search schedule with 5 searches (including a search at t = 0), and the split_SS variable would have values for each of the 4 search intervals. The levels of the split_SS must be in contiguous blocks. For example, season = c("S", "S", "F", "F") would be acceptable, but season = c("S", "F", "S", "F") would not be.

split_time variables are numeric vectors that split the monitoring period into distinct time intervals. For example, split_time = c(0, 30, 60, 90, 120) would split the 120 monitoring period into 30-day intervals, and calcSplits() would return mortality estimates for each of the intervals.

Value

An object of class splitFull is returned. If one splitting covariate is given, then the output will be an array of estimated mortality in each level of the splitting covariate, with one row for each covariate level and one column for each simulation draw. If two splitting covariates are given, output will be a list of arrays. Each array gives the estimated mortalities for one level of the second splitting covariate and all levels of the first splitting covariate.

Objects of class splitFull have attributes vars (which gives the name of the splitting covariate(s)) and type (which specifies whether the covariate(s) are of type split_CO, split_SS, or split_time). A summary of a resulting splitFull object is returned from the S3 function summary(splits, CL = 0.90, ...), which gives the mean and a 5-number summary for each level of each covariate. The 5-number summary includes the alpha/2, 0.25, 0.5, 0.75, and 1 - alpha/2 quantiles, where alpha = 1 - CL. A graph summarizing the results can be drawn using plot(splits, CL, ...), which gives a graphical representation of the summary.

Examples

 
  model_SE <- pkm(p ~ 1, k ~ 1, data = wind_RPbat$SE)
  model_CP <- cpm(l ~ 1, s ~ 1, data = wind_RPbat$CP, dist = "weibull",
    left = "LastPresent", right = "FirstAbsent")
  Mhat <- estM(nsim = 1000, data_CO = wind_RPbat$CO, 
    data_SS = wind_RPbat$SS, data_DWP = wind_RPbat$DWP, 
    model_SE = model_SE, model_CP = model_CP,
    unitCol = "Turbine", COdate = "DateFound")

  M_spp <- calcSplits(M = Mhat, split_CO = "Species",
    data_CO = wind_RPbat$CO)
  summary(M_spp)
  plot(M_spp)