We apply an exploratory spatial data analysis framework for integrating the time series of hedge fund returns to its neighborhood, mapping, and local analysis for the feasible spatial modeling. Our approach takes into account option-like features and serial correlations in the stylized hedge funds' risk-return payoffs. By comparing the classic risk factor analysis of hedge fund performance of ordinary least squares regression with spatial autoregressive models, we investigate each model’s respective ability to estimate the stylized risk premiums. The time series analysis of hedge fund returns from the Barclays Hedge indicates that, for some of the sub-investment styles such as equity long-short, equity long-bias, event-driven arbitrage, convertible arbitrage, fixed-income arbitrage, distressed securities, multi-strategies, and commodity trading advisors, the spatial autoregressive modeling may provide consistent estimates of factor risk-premiums by correcting structural spatial dependence through the measure of endogeneity of implied volatilities. We employ spatial specifications including spatial lag (SLM) and spatial error (SEM) models to minimize the overestimation bias in factor risk premiums by exploring some practical implications in an ad hoc screening through the missing spatial autoregressive heterogeneity in the ordinary least squares approach. Both SLM and SEM models are applied to verify a ‘meant-to-be’ spatial dependence to a relatively short time series of a recently failed credit hedge fund previously marketed its vanishingly rare talent of return predictability and consistency.