Local and remote wind-coherent responses of sea surface heights (SSHs) off the US West Coast (USWC) are described with statistical and analytical models. The wind transfer functions are statistically derived from surface wind stress at National Data Buoy Center (NDBC) buoys, located within 50 km from the shoreline, and detided SSHs (SSH anomalies; SSHAs) at shoreline tide gauges for 15 years (1995 to 2009) using linear regression in the frequency domain. A two-dimensional analytical model constrained by the coastal boundary provides a dynamical framework to interpret the data-derived statistical model. Although both transfer functions agree well at low frequency [sigma a parts per thousand currency sign 0.4 cycles per day (cpd)], they appear to be inconsistent at high frequency (sigma a parts per thousand yen 0.8 cpd; e.g., diurnal and its harmonic frequencies) because of incoherent signals between wind stress and SSHAs as well as their low signal-to-noise ratios. A multivariate regression analysis using wind stress at multiple wind buoys is implemented with a modified expectation maximization. The cross-validated skill increases and becomes saturated as the number of regression basis functions increases, demonstrating the influence of local and remote winds. The skill computed from all available winds off the USWC has a maximum as 0.1 in southern California, 0.2 to 0.3 in central California, and 0.3 to 0.5 in northern California, Oregon, and Washington. The residual SSHAs, incoherent components with all available coastal wind stress off the USWC, still contain poleward propagating signals, considered as components forced by remote winds outside of the domain.