Assessing Bias in Citizen Science Data through Linear Regression

Citizen Science Data is a valuable resource for addressing ecological questions. Some initiatives like eBird have gained prominence since its establishment in 2002. This platform has emerged as a vital source for collecting birdwatching data, particularly beneficial for ornithology and ecology. However, the semi-structured nature of this data introduces potential biases during the collection process. In ecology, bias—especially in the context of citizen or community science—refers to a systematic error that can distort data collection and analysis, potentially leading to inaccurate conclusions about ecological relationships. These ecological biases include taxonomic, spatial, and temporal factors. While established preprocessing techniques can mitigate some of these issues, specific covariates necessitate more refined treatments to ensure the reliability of the conclusions drawn from this data. This study introduces a Linear Regression-based hypothesis testing technique from correlation measures to evaluate bias in Citizen Science Data corresponding to sightings of the Monk Parakeet (Myiopsitta monachus). The results demonstrate that this straightforward yet potent model can capture sufficient variance to eliminate effort variables and preserve other covariates for future research. The proposed approach leverages the observed significance and the coefficient of determination to assess performance.