In this research, the one of the most important model and widely used in many and applications is linear mixed model, which widely used to analysis the longitudinal data that characterized by the repeated measures form .where estimating linear mixed model by using two methods (parametric and nonparametric) and used to estimate the conditional mean and marginal mean in linear mixed model ,A comparison between number of models is made to get the best model that will represent the mean wind speed in Iraq.The application is concerned with 8 meteorological stations in Iraq that we selected randomly and then we take a monthly data about wind speed over ten years Then average it over each month in corresponding year, so we get different clusters ,each cluster contain 12 observation that represent a mean wind speed for each station . The comparison among the best models are held by using statistical standard the mean square Error(MSE),our conclusion for the parametric model during the application the with additional random effect(the second model) is better than the model without addithonal random effect(the first model)for all station in general,for nonparametric model we found the conditional local mixed model is better than marginal mixed model in estimation the conditional and marginal means for mixed model in general, for marginal mean , where found that the marginal local mixed model is better for all the stations that we were sampled except for the fifth station we found that the conditional local mixed model is better for the marginal local mixed model in estimation of marginal mean mixed model .
