format_data {popbayes}R Documentation

Format count series

Description

This function provides an easy way to get count series ready to be analyzed by the package popbayes. It must be used prior to all other functions.

This function formats the count series (passed through the argument data) by selecting and renaming columns, checking columns format and content, and removing missing data (if na_rm = TRUE). It converts the original data frame into a list of count series that will be analyzed later by the function fit_trend() to estimate population trends.

To be usable for the estimation of population trends, counts must be accompanied by information on precision. The population trend model requires a 95% confident interval (CI). If estimates are total counts or guesstimates, this function will construct boundaries of the 95% CI by applying the rules set out in https://frbcesab.github.io/popbayes/articles/popbayes.html. If counts were estimated by a sampling method the user needs to specify a measure of precision. Precision is preferably provided in the form of a 95% CI by means of two fields: lower_ci and upper_ci. It may also be given in the form of a standard deviation (sd), a variance (var), or a coefficient of variation (cv). If the fields lower_ci and upper_ci are both absent (or NA), fields sd, var, and cv are examined in this order. When one is found valid (no missing value), a 95% CI is derived assuming a normal distribution. The field stat_method must be present in data to indicate if counts are total counts ('T'), sampling ('S'), or guesstimate ('X').

If a series mixes aerial and ground counts, a field field_method must also be present and must contain either 'A' (aerial counts), or 'G' (ground counts). As all counts must eventually refer to the same field method for a correct estimation of trend, a conversion will be performed to homogenize counts. This conversion is based on a preferred field method and a conversion factor both specific to a species/category. The preferred field method specifies the conversion direction. The conversion factor is the multiplicative factor that must be applied to an aerial count to get an equivalent ground count (note that if the preferred field method is 'A', ground counts will be divided by the conversion factor to get the equivalent aerial count).

The argument rmax represents the maximum change in log population size between two dates (i.e. the relative rate of increase). It will be used by fit_trend() but must be provided in this function.

These three parameters, named pref_field_method, conversion_A2G, and rmax can be present in data or in a second data.frame (passed through the argument info). Alternatively, the package popbayes provides their values for some African large mammals.

Note: If the field field_method is absent in data, counts are assumed to be obtained with one field method.

Usage

format_data(
  data,
  info = NULL,
  date = "date",
  count = "count",
  location = "location",
  species = "species",
  stat_method = "stat_method",
  lower_ci = "lower_ci",
  upper_ci = "upper_ci",
  sd = NULL,
  var = NULL,
  cv = NULL,
  field_method = NULL,
  pref_field_method = NULL,
  conversion_A2G = NULL,
  rmax = NULL,
  path = ".",
  na_rm = FALSE
)

Arguments

data

a data.frame with at least five columns: location, species, date, count, and stat_method.

The stat_method field indicates the method used to estimate counts. It can contain: T (total counts), X (guesstimate), and/or S (sampling).

If individual counts were estimated by sampling, additional column(s) providing a measure of precision is also required (e.g. lower_ci and upper_ci, or sd, cv, var). Precision metrics can be different between counts. For instance, some sampling counts can have a sd value and others lower_ci and upper_ci. In that case three columns are required (lower_ci, upper_ci, and sd). See above section Description for further information on the computation of the 95% confident interval of estimates.

If the individuals were counted by different methods, an additional field field_method is also required. It can contain: G (ground counts) and/or A (aerial counts). See above section Description for further information on the counts conversion.

Others fields can be present either in data or info (see below).

info

(optional) a data.frame with species in rows and the following columns: species (species name), pref_field_method, conversion_A2G, and rmax. See above section Description for further information on these fields. Default is NULL (i.e. these information must be present in data if not available in popbayes).

date

a character string. The column name in data of the date. This column date must be in a numerical form with possibly a decimal part. Default is 'date'.

count

a character string. The column name in data of the number of individuals. This column must be numerical. Default is 'count'.

location

a character string. The column name in data of the site. This field is used to distinguish count series from different sites (if required) and to create an unique series name. Default is 'location'.

species

a character string. The column name in data (and in info if provided) of the species. This field is used to distinguish count series for different species (if required) and to create an unique series name. Default is 'species'.

stat_method

a character string. The column name in data of the method used to estimate individuals counts. It can contain 'T' (total counts), 'X' (guesstimate), and/or 'S' (sampling). If some counts are coded as 'S', precision column(s) must also be provided (see below). Default is 'stat_method'.

lower_ci

(optional) a character string. The column name in data of the lower boundary of the 95% CI of the estimate (i.e. count). If provided, the upper boundary of the 95% CI (argument upper_ci) must be also provided. This argument is only required if some counts have been estimated by a sampling method. But user may prefer use other precision measures, e.g. standard deviation (argument sd), variance (argument var), or coefficient of variation (argument cv). Default is 'lower_ci'.

upper_ci

(optional) a character string. The column name in data of the upper boundary of the 95% CI of the estimate (i.e. count). If provided, the lower boundary of the 95% CI (argument lower_ci) must be also provided. Default is 'upper_ci'.

sd

(optional) a character string. The column name in data of the standard deviation of the estimate. Default is NULL.

var

(optional) a character string. The column name in data of the variance of the estimate. Default is NULL.

cv

(optional) a character string. The column name in data of the coefficient of variation of the estimate. Default is NULL.

field_method

(optional) a character string. The column name in data of the field method used to count individuals. Counts can be ground counts (coded as 'G') or aerial counts (coded as 'A'). This argument is optional if individuals have been counted by the same method. See above section Description for further information on the count conversion. Default is NULL.

pref_field_method

(optional) a character string. The column name in data of the preferred field method of the species. This argument is only required is field_method is not NULL (i.e. individuals have been counted by different methods). Alternatively, this value can be passed in info (or internally retrieved if the species is listed in the package). See above section Description for further information on the count conversion. Default is NULL.

conversion_A2G

(optional) a character string. The column name in data of the count conversion factor of the species. This argument is only required if field_method is not NULL (i.e. individuals have been counted by different methods). Alternatively this value can be passed in info (or internally retrieved if the species is listed in the package). See above section Description for further information on the count conversion. Default is NULL.

rmax

(optional) a character string. The column name in data of the species demographic potential (i.e. the relative rate of increase of the population). This is the change in log population size between two dates and will be used later by fit_trend(). Default is NULL.

path

a character string. The directory to save formatted data. This directory must exist and can be an absolute or a relative path. Default is the current working directory.

na_rm

a logical. If TRUE, counts with NA values will be removed. Default is FALSE (returns an error to inform user if NA are detected).

Value

An n-elements list (where n is the number of count series). The name of each element of this list is a combination of location and species. Each element of the list is a list with the following content:

Note: Some original series can be discarded if one of these two conditions is met: 1) the series contains only zero counts, and 2) the series contains only a few dates (< 4 dates).

Examples

## Load Garamba raw dataset ----
file_path <- system.file("extdata", "garamba_survey.csv", 
                         package = "popbayes")
                         
garamba <- read.csv(file = file_path)

## Create temporary folder ----
temp_path <- tempdir()

## Format dataset ----
garamba_formatted <- popbayes::format_data(
  data              = garamba, 
  path              = temp_path,
  field_method      = "field_method",
  pref_field_method = "pref_field_method",
  conversion_A2G    = "conversion_A2G",
  rmax              = "rmax")

## Number of count series ----
length(garamba_formatted)

## Retrieve count series names ----
popbayes::list_series(path = temp_path)

## Print content of the first count series ----
names(garamba_formatted[[1]])

## Print original data ----
garamba_formatted[[1]]$"data_original"

## Print converted data ----
garamba_formatted[[1]]$"data_converted"

[Package popbayes version 1.2.0 Index]