中文摘要：

石制品组合间对比是石制品研究的关键一环,定量分析凭借其在复杂变量对比研究中具有的客观优势拓宽了石制品组合间对比的视野,与考古学思想、石制品组合内涵解读的深化具有相辅相成的关系。一般而言,定量分析方法分为图表法和统计法两类,主要包括三个基本步骤：选择对比组合,确定研究目标;根据研究目标提取变量;选择合适的定量分析方法。值得注意的是,考古学解释仍为石制品研究的重点,定量分析方法更多作为一种辅助手段,应用时需检验其有效性。

英文摘要：

Comparing assemblages is a key to lithic analysis. From the history perspective, the interpretation of lithic assemblage may be outlined two aspects：l）a lithic assemblage is sets of objects and features generally from a site or a layer, we might get the meaning of technology, culture, or even function from the analysis of presence or absence of its material or composition; 2）lithic assemblage was formed in a complex process, therefore, besides concerning similarity and difference between the assemblages, factors like the boundary of the assemblage and the life cycle of artifacts should also be taken into consideration. Nowadays, under the circumstance of interdiscipline, with the objective advantages in multivariate comparative study, quantitative analysis broadens the horizon of comparing assemblages. The method includes graphical method and statistics. However, such kind of practice is relatively insufficient in China. The advantages of graphical method are those could present complex data multidimensionally and simultaneously and also contribute to increase the readability. It chiefly includes Cumulative Graph/Curves, Clustered Histogram/Accumulative Histogram, Snowflake Diagram and Radar Graph. The characteristic of the latter two is that they could reduce the variates with classification, and obtain more information. Nevertheless, it is better to combine the methods like use-wear analysis and residue analysis in classification to improve the accuracy of the information. In statistics, assemblage diversity has been paid close attention by researchers all the time, including the aspects of richness, evenness and heterogeneity. The available methods could be regression method, simulation method, Shannon-Weaver information statistic H, Pielou statistic J and so on. Considering the sample effect, researchers now prefer nonparametric estimators like Jackknife Technique, Bootstrap, Rarefaction （Interpolation） and Extrapolation. With the correlation and confidential intervals given by these methods, resear