For a given eBird Status and Trends species, produce a line plot showing the partial dependence (PD) relationship for a given predictor. Two options for smoothing are provided.
plot_pds(
pds,
predictor,
ext,
bootstrap_smooth = TRUE,
show_stixel_pds = FALSE,
show_quantiles = FALSE,
n_bs = 100,
ss_equivalent = 10,
k = 25,
ci_alpha = 0.05,
gbm_n_trees = 500,
ylim = NULL,
plot = TRUE
)data frame; partial dependence data from load_pds().
character; single predictor name to plot PD for. For a full list of predictors, and their associated definitions, see ebirdst_predictors.
ebirdst_extent object; the spatiotemporal extent over which to calculate PDs. This is required, since results are less meaningful over large spatiotemporal extents.
logical; the ideal visualization of the PD data is a
pointwise GAM smoothing of the individual stixel PD values. This argument
specifies whether this should be done directly on the full PD dataset
(bootstrap_smooth = FALSE) or by subsampling and bootstrapping. The
latter approach deals with the randomness in the data and can be more
efficient for large datasets.
logical; whether to plot the individual stixel PD values as semi-transparent lines.
logical; adds a band for the upper (90th) and lower (10th) quantiles of the individual stixel PD values. These are calculated using quantile regression.
int; number of GAM bootstrap iterations when estimating PD
confidence intervals. Ignored if bootstrap_smooth = FALSE.
int; when bootstrapping to estimate PD confidence
intervals, this argument specifies the size of the subsample of the
original data. In particular, ss_equivalent should be an integer
representing the equivalent sampling size when averaging this number of PD
estimates.
integer; number of knots to use in the GAM when smooth the PD relationship.
numeric; alpha level of confidence intervals. Default is 0.05.
integer; number of trees to fit in the GBM when estimating
quantiles. Ignored if show_quantiles = FALSE. Default is 500.
numeric; 2-element vector to pre-define the y-limits of plotting.
In the format c(ymin, ymax).
logical; whether to plot the PD relationships or just return data.
Plots the smoothed partial dependence relationship for the specified predictor and returns a data frame of the smoothed curve with confidence intervals.
if (FALSE) {
# download example data
path <- ebirdst_download("example_data", tifs_only = FALSE)
# or get the path if you already have the data downloaded
path <- get_species_path("example_data")
# load predictor dependence data
pds <- load_pds(path)
# define a spatiotemporal extent to plot data from
bb_vec <- c(xmin = -86, xmax = -83, ymin = 41.5, ymax = 43.5)
e <- ebirdst_extent(bb_vec, t = c("05-01", "05-31"))
# for testing, run with 5 bootstrap iterations for speed
# in practice, best to run with the default number of iterations (100)
pd_smooth <- plot_pds(pds, "solar_noon_diff", ext = e, n_bs = 5)
dplyr::glimpse(pd_smooth)
}