proxydata

Proxy data are measurements or observations used to approximate a target quantity when direct observation is difficult or impossible. By definition, a proxy is correlated with the variable of interest and is assumed to track its changes under a defined context. Proxy data are widely used across disciplines, including climate science, economics, epidemiology, and social science.

In climate research, proxy records—such as tree-ring widths, ice-core isotopes, and sediment compositions—reconstruct past temperatures, precipitation, and environmental conditions when instrumental data are unavailable. In economics and public health, proxies include indicators like consumer sentiment, electricity consumption, search engine query volumes, or pharmaceutical sales to estimate current demand, disease incidence, or well-being.

Constructing and using proxy data involves selecting proxies with a plausible link to the target variable, calibrating the relationship against observed data, and validating the model with out-of-sample tests. Methods range from simple regression and time-series models to more complex latent-variable or Bayesian approaches that account for measurement error and uncertainty. Calibration is critical to avoid bias and overfitting, and researchers must assess stability of the proxy-target relationship over time and across contexts.

Limitations include measurement error, nonstationarity, and potential biases if the proxy is weakly related or responds to unrelated influences. Transparent reporting of proxy selection, assumptions, and uncertainty is essential to credible inference.

See also: proxy variable, latent variable, calibration, instrumental variable.