Accuracy of wind speed data has important impact on determining wind power output from a wind turbine. There are many researches on four widely used wind speed distribution models described by gamma, lognormal, Rayleigh and Weibull for assessing wind potentials. However, there is lack of studies to evaluate sensitivity of these models with respect to accuracy of the measured wind data. In this paper, wind speed data are measured by national data buoy center (NDBC) over ten years, from 2004 to 2014, for four offshore stations in the east of the U.S. Two methods of maximum-likelihood estimator (MLE) and method of moments (MOM) are utilized for calculating parameters involved with these four distribution functions. For reducing the accuracy, a truncated set of wind data is generated by removing the decimal digits of the wind data; reducing the resolution to 1 m/s. Also, the best distribution functions in terms of performance are selected by examining nine goodness-of-fit statistics. From the outcomes, it is concluded that the Weibull function offers a better fit to the both actual and truncated data. Additionally, the Rayleigh distribution function exhibits suitable fit with the truncated wind speed data.