中文摘要：

目的采用3种常用统计学方法构建上海市社区成年居民膳食模式,并鉴定与超重肥胖有关的膳食模式。方法对上海市2012年冬至2014年秋膳食调查中,2 812名研究个体4个季度4 d的24 h膳食回顾调查数据用聚类分析、因子分析以及综合聚类分析和因子分析结合法来确定主要的膳食模式,用logistic回归分析讨论其与超重肥胖之间的关系。结果 logistic回归分析显示,控制年龄、性别、文化程度、运动情况、能量等混杂因素后,因子分析中水果饮料型膳食模式相对于传统型膳食模式显示较高的超重风险（OR=1.25,P=0.04）,高蛋白型膳食模式相对于主食型膳食模式显示较低的肥胖风险（OR=0.77,P=0.01）;而采用综合聚类分析和因子分析的结合法则显示,传统型膳食模式相对于均衡膳食模式显示较低的肥胖风险（OR=0.82,P=0.04）。结论聚类分析、因子分析,以及两者结合法均可用来分析膳食模式,为进一步鉴定疾病相关膳食模式提供帮助,因子分析和结合法比聚类分析则相对更优。

英文摘要：

Objective To build the dietary patterns in Shanghai adults by applying 3 statistical methods,and identify the overweight-related or obesity-related dietary patterns. Methods 24 hours dietaryrecall questionnaire was used to collect the dietary data of 2812 adults in Shanghai from the winter of2012 to the autumn of 2014. Cluster analysis,factor analysis and the combination of the above 2 methods were used to identify the dietary patterns; logistic regression was used to analysis the relationship between dietary patterns and being overweight or obesity. Results After controlling the confounding facts of age,gender,degree of education,physical activity and energy,the logistic regression analysis showed that there was higher risk of being overweight in ＂fruits ＆ beverage＂pattern than ＂traditional food＂pattern（ OR =1. 25,P = 0. 04） and lower risk of being overweight in ＂high protein＂pattern than ＂staple food＂pattern（ OR = 0. 77,P = 0. 01）. Combination of cluster analysis and factor analysis showed that there was lower risk of being obesity in ＂traditional food＂pattern than ＂balanced＂pattern（ OR = 0. 82,P = 0. 04）.Conclusion Cluster analysis,factor analysis,and the combination method can be used to analysis dietary patterns for the further exploration of their relationships with obesity,though factor analysis and thecombination method is doing better than cluster analysis.