Factor amp; Cluster Analyses因子amp；聚类分析.pptVIP

Factor Cluster Analyses Factor Analysis Goals Data Process Results Factor Analysis Goals Reduce number of variables Typically for further analysis Measure/Understand underlying construct e.g. What is intelligence, beauty, effectiveness? Factor Analysis Data Typically numeric Variables must have some intercorrelations Factor Analysis Data Factor Analysis Process Linear combination of variables Types: Principal Components Analysis Factor Analysis Maximize variance in each Factor / Component, with 0 covariance between components. Factor Analysis Results Cluster Analysis Goal Data Process Results Cluster Analysis Goals Find subgroups within a larger group Create profiles of subgroups for further action (marketing, medical intervention, etc.) Cluster Analysis Data Typically numeric Free of correlations Free of outliers Cluster Analysis Data Cluster Analysis Process K-means clustering Prespecified number of clusters Based on Euclidean distances Hierarchical Tree Each observation is a cluster, and the number of clusters is iteratively reduced Cluster Analysis Results Cluster Means The mean of each variable for all the observations within the cluster is output. The combined set of these means for each cluster is called the cluster Centroid. Cluster Analysis Results * * These numbers show relative importance of each variable within a component *