Current issue: 54(2)
Under compilation: 54(3)
In this study, model-based and design-based inference methods are used for estimating mean volume and its standard error for systematic cluster sampling. Results obtained with models are compared to results obtained with classical methods. The data are from the Finnish National Forest Inventory. The variation of volume in ten forestry board districts in Southern Finland is studied. The variation is divided into two components: trend and correlated random errors. The effect of the trend and the covariance structure on the obtained mean volume and standard error estimates is discussed. The larger the coefficient of determination of the trend model, the smaller the model-based estimates of standard error, when compared to classical estimates. On the other hand, the wider the range and level of autocorrelation between the sample plots, the larger the model-based estimates of standard error.