Abstract: The purposive sampling method based on volume arithmetic mean of typical subcompartment (PSMAMT) is a new method of forest inventory at present in China. It is established on the basis of the approximate information of forest resource which has already been achieved. According to the forest resource files and remote sensing information, all subcompartments in a population are divided into several types. Only a few samples which are sampled based on volume arithmetic mean of typical subcompartment are investigated, and the population parameters with considerable accuracy can be made thereupon.PSMAMT is noted for its use of 3 P-sampling (probability proportional to prediction sampling). With 3 P sample trees, the estimation of population volume, increment or its percent, timber assortment volume or its percent can be carried out to eliminate the error caused by the disagreement between the table applied and the investigation population. The experiment of PSSAMT was conducted by various methods of investigation such as point sampling, linear sampling and strip sampling, and the data of 710 sample trees were collected, from which the following three main conclusions have been drawn.1 . The standing volume of each sample tree from the three samplings was measured by height accumulation method and Pressler method to estimate the population volume. At the same time, one-way volume table, general volume table and standard table were used to calculate the stand volume. Results from all the sixteen different methods show that the accuracy is more than 85%. 2 . The mean increment percent of population was estimated according to different types and species of trees by single tree increment percent method, with the B. H. diameter increment of 3 P sample trees being measured. The accuracy is more than 90%.3 . The table of merchantable volume percent has been worked out from the information of 3 P sample trees by standing tree diameter difference method, and the results were satisfying when it was compared with that from the mean sample trees method.