Plot the evolution of a diving parameter accross time between two populations

This function approximates the relationship between the considered diving behavior and time in order to better represent this evolution by a smooth curve, rather than by a scatterplot.

Usage

plot_comp(
  data,
  diving_parameter = NULL,
  group_to_compare = "sp",
  time = "day_departure",
  id = ".id",
  nb_days = 100,
  bs = "cs",
  k = 6,
  alpha_point = 0.01,
  alpha_ribbon = 0.4,
  linetype_ribbon = 2,
  colours = NULL,
  ribbon = TRUE,
  point = TRUE,
  individual = TRUE,
  populational = TRUE,
  rows = NULL,
  cols = NULL,
  scales = "fixed",
  method = "REML",
  export_data_model = FALSE
)

Arguments

data: A dataset containing the required information
diving_parameter: The colname associated with the diving parameter to represent
group_to_compare: The colname associated with the groups to compare
time: The colname associated with time
id: The colname associated with individual
nb_days: How many days to represent
bs: Smooth terms in GAM
k: The dimension of the basis used to represent the smooth term
alpha_point: The transparency of the point
alpha_ribbon: The transparency of the ribbon
linetype_ribbon: Line type for ribbon border
colours: The colours to use
ribbon: Should confidence interval be added
point: Should the points be displayed
individual: Should individuals curves be displayed
populational: Should populational curve be displayed
rows: The colname used for a facet in row
cols: The colname used for a facet in column
scales: Are scales shared across all facets (the default, "fixed")
method: The smoothing parameter estimation method for the GAM (default REML)
export_data_model: Boolean to export the data of the underlying model

Value

Return a ggplot

Details

This function fits a GAM with the species as a grouping factor and a random effect (intercept + slope) on the individual (i.e. diving parameter ~ species + s(time) + (1 time | individual). This allows to represent a curve for each species, but also to access to the curve associated with each individual.

References

https://stats.stackexchange.com/questions/403772/different-ways-of-modelling-interactions-between-continuous-and-categorical-pred

https://github.com/DistanceDevelopment/dsm/wiki/Why-is-the-default-smoothing-method-%22REML%22-rather-than-%22GCV.Cp%22%3F

James Grecian. (2022). jamesgrecian/harpPup: v1.0 (v1.0). Zenodo. https://doi.org/10.5281/zenodo.5901391

Examples

if (FALSE) {
# load data
data_nes <- get_data("nes")
data_ses <- get_data("ses")

# combine
data_comp <- rbind(
  rbindlist(data_nes$year_2018),
  rbindlist(data_ses$year_2014),
  use.names = T,
  fill = T
)

# plot
plot_comp(data_comp, "maxdepth")
}