This function runs a PCA (Principal component analysis)
or a t-SNE (t-distributed Stochastic Neighbor Embedding) over the
global and pairwise metrics of collective motion per each event to
produce a swarm space. The PCA is computed with the stats::prcomp
function and the t-SNE with the Rtsne::Rtsne function.
Usage
swarm_space(
metrics_data,
space_type = "pca",
event_dur_limit = NA,
tsne_rand_seed = NA,
tsne_perplexity = 25
)Arguments
- metrics_data
A dataframe with metrics of collective motion per event.
- space_type
A string, stating the choice between PCA ("pca") and t-SNE ("tsne"), default = "pca".
- event_dur_limit
Numeric, capturing an event duration value in seconds. Used to filter out events that are shorter that this value. Default = NA, no filtering is applied.
- tsne_rand_seed
Numeric, the random seed for the t-SNE analysis, to ensure reproducibility. Default = NA, but a value should be given if the t-SNE analysis is selected.
- tsne_perplexity
Numeric, the perplexity parameter for the t-SNE analysis. Usually between 10-50, default = 25.
Value
A list with 3 elements: a dataframe representing the
swarm space (x and y coordinates per event of each species), a reference dataframe
(ref) including all the additional event information from the input metric data
dataframe, a dataframe for the t-SNE analysis (tsne_setup) that includes the
input parameters used, and a list for the PCA analysis (pca) with the output
of the stats::prcomp command.
Author
Marina Papadopoulou m.papadopoulou.rug@gmail.com
Examples
data(multi_species_metrics)
ss <- swarm_space(multi_species_metrics)