Effects of ignoring survey design information for data reuse

Area of Science:

• Ecology
• Ecological research
• Data science

Background:

• Ecological research increasingly relies on data reuse from aggregated databases.
• Ensuring effective and unbiased data reuse requires assessing the adequacy of information provided by these databases.
• Survey designs with uneven inclusion probabilities (e.g., stratified sampling) are common in ecological data collection.

Purpose of the Study:

• To investigate whether aggregated databases provide sufficient information for effective and unbiased reuse of ecological data.
• To evaluate the impact of ignoring survey design information on ecological density estimates.
• To identify methods for mitigating estimation bias in ecological data analysis.

Main Methods:

• A simulation experiment was conducted using datasets with varying degrees of uneven inclusion probabilities.
• The study examined the resulting estimates of population density (individuals per unit area).
• The impact of using naive analytical methods versus methods incorporating design information was assessed.

Main Results:

• Ignoring survey design information can lead to profound biases in density estimates, reaching up to 250%.
• Increasing data volume does not inherently reduce this density estimation bias.
• Estimation bias can be significantly mitigated by employing appropriate estimators or models that utilize survey design information.

Conclusions:

• Essential survey design information, including sample location selection processes and covariates, must be stored and served with ecological data.
• Availability of design information is critical for enabling meaningful inference and supporting robust data reuse.
• Incomplete metadata in aggregated databases hinders accurate ecological analysis and data reuse.