EARLY HEARING DETECTION AND INTERVENTION VIRTUAL CONFERENCE
MARCH 2-5, 2021
(Virtually the same conference, without elevators, airplane tickets, or hotel room keys)
5/25/2021 | 2:00 PM - 3:30 PM | Minding the data gap: Comparing probabilistic approaches to estimate missing data in time-series | Virtual Platform
Minding the data gap: Comparing probabilistic approaches to estimate missing data in time-series
Missing data are common in research, and are particularly problematic when observations are auto-correlated in time. In aquatic ecosystems, sensors collect continuous data, but these data are prone to missingness due to sensor malfunction. It is unclear how missing data gaps affect parameter estimation in time-series models. We tested the effectiveness of two probabilistic approaches, multiple imputations (MI) and Bayesian parameter estimation (Bayes), to estimate missing data and recover known parameters of a time-series model of simulated light-controlled gross primary production. The MI approach imputes missing data points five times, generating five complete datasets, with each dataset modeled individually and inference made on model-averaged parameters. The Bayes approach lists missing data as parameters to be estimated in a Bayesian time-series model. Both approaches effectively estimated a range of randomly removed data (1-40% of the dataset) in single-day and weeklong blocks. Furthermore, both approaches returned the known time-series model parameters, and more importantly, parameter estimation was not affected by varying the amount of missing data (up to 40%). Probabilistic approaches to estimating missing data may be useful to fill data gaps without sacrificing the accuracy of parameter estimation for simple models.
- Models
- Big data
- Biogeochemistry
Presentation:
This presentation has not yet been uploaded.
Handouts:
Handout is not Available
Transcripts:
CART transcripts are NOT YET available, but will be posted shortly after the conference
Presenters/Authors
Matt Trentman
(), Flathead Lake Biological Station, University of Montana, matt.trentman@flbs.umt.edu;
ASHA DISCLOSURE:
Financial -
Nonfinancial -
Robert O. Hall
(), Flathead Lake Biological Station, University of Montana, bob.hall@flbs.umt.edu;
ASHA DISCLOSURE:
Financial -
Nonfinancial -
Laurel Genzoli
(), University of Montana, laurel.genzoli@umontana.edu;
ASHA DISCLOSURE:
Financial -
Nonfinancial -