Supplementary information (data) for: A comparison of statistical methods for deriving occupancy estimates from acoustic monitoring data.

Dataset used to run the occupancy models described in Lydia K.D. Katsis, Tessa A. Rhinehart, Elizabeth Dorgay, Emma E. Sanchez, C. Patrick Doncaster, Jake L. Snaddon, Justin Kitzes. A comparison of statistical methods for deriving occupancy estimates from acoustic monitoring data.

Data consists of
Ground-truth
1) ground_truth_data.csv: csv of manually reviewed files for presence or absence of howler monkeys, used as benchmark to compare modelled occupany estimates.

Partially annotated datasets
1) top_ten_annotated.csv : files with top-10 machine learning scores per site manually reviewed (highest scoring file on every third day)
2) random_ten_annotated.csv : 10 randomly selected files per site manually reviewed (randomly selected file on every third day)
3) scheduled_listening_annotated.csv: first file after 5am every 3 days from each site manually reviewed.

Machine learning predictions on temporally subset dataset
1) dawn_all.csv : full dataset of predictions, with no subsetting
2) dawn_10_max.csv : temporal interval of 10 minutes, with file with maximum machine learning score sampled
3) dawn_10_random.csv: temporal interval of 10 minutes, first file sampled
4) dawn_30_max.csv: temporal interval of 30 minutes, with file with maximum machine learning score sampled
5) dawn_30_random.csv: temporal interval of 30 minutes, first file sampled

Columns in data consist of:
LocationID/site: unique identifier for recorder locationdate/timestamp: when the recording was madeannotation: manualy review for presence (1) or absence (0) of howler monkey in cliplogit_present: machine learning score with logit transformationsoftmax_present: machine learning score with softmax transformation