This study addresses possible obstacles and errors which can be expect
ed when one applies the autoregressive method for simulation of wind h
istories. The mean value, variance, and power spectra of the wind spee
d of the generated sample are assessed by comparison with the target p
arameters required. The influence of the order of autoregression and o
f the number of sample points is investigated. A confidence interval p
robability is used as a measure of the accuracy of the simulation. Num
erical results obtained from the example of the wind load on an antenn
a mast are provided.