A model-based solution for observational errors in laboratory studies.

Area of Science:

• Ecology
• Molecular Biology
• Biostatistics

Background:

• Molecular detection methods for various organisms and agents are prone to false-negative and false-positive errors.
• These errors can lead to inaccurate conclusions regarding occurrence, prevalence, sensitivity, specificity, and relationships with covariates.
• Existing occupancy models primarily address false-negative errors and have limitations in handling false-positive errors.

Purpose of the Study:

• To extend the occupancy modelling framework to account for false-positive errors in both negative and positive samples.
• To provide unbiased inference on occurrence and detection probabilities.
• To enable reliable conclusions about the effectiveness of sampling designs, handling protocols, and diagnostic tests.

Main Methods:

• Developed an extended occupancy modelling framework incorporating false-positive error rates for both negative and positive samples.
• Applied the model to simulated data to assess its parameter recovery and performance compared to existing methods.
• Utilized the model on an experimental dataset for Batrachochytrium dendrobatidis detection in amphibians.

Main Results:

• The extended model accurately recovered known parameters in simulated data.
• It demonstrated superior performance over common approaches when dealing with observational errors.
• Application to Batrachochytrium dendrobatidis data yielded reliable inference despite data imperfections.

Conclusions:

• The proposed occupancy model effectively corrects for both false-negative and false-positive errors in molecular detection data.
• This approach yields unbiased inference on ecological parameters and diagnostic test performance.
• It eliminates the need for arbitrary thresholds, offering more robust and reliable scientific conclusions.