多变量统计概说IntroductiontoMultivariateStatistics.PPTVIP

多變量統計概說Introduction to Multivariate Statistics 多變量統計…… 我們在哪裡？ Where are we? 1. 原理概念的深化 2. 技術的複雜化 3. 研究設計的複雜化 統計原理與技術的關係 量化研究與多變量統計 多變量統計：基本概念 Why multivariate? Conceptual complexity vs. technical complexity Design driven vs. Analysis driven Definition Analysis of multiple IVs and DVs From uni-variate to bi-variate to multi-variate Research Design and Multivariate Experimental? When Manipulation of IV and measurement of DV Random assignment of subject Controlling for the influential factors Confirmation of Causality Challenges to Experimental Design Statistical treatment on multiple/intercorrelated DVs Inflation of the type I error Non-experimental Design Relationship of various dependent variables Low sensitivity of the univariate statistics Measurement of multivariate data Measurement: 運用一套符號系統去描述某個被觀察對象的某個屬性（attribute）的過程（Nunnallly ＆ Bernstein，1994） 利用不同的量尺，對於現象加以『變項化』測定的過程是為測量 Type of variables Categorical/Discrete variable (nominal and ordinal) Continuous variable (interval and ordinal) Special form of variables Ordinal variable Dummy variable From discrete to dichotomous From continuous to categorical 測量量尺的數學關係 Population and sampling Statistical purpose: inferential from sample to population Basic requirement of the sample In non-experimental design: Representative Random sampling Relative bigger size In experimental design Homogeneous Random assignment Relative smaller size Analysis to multivariate data Orthogonality Independence of the variables Application addition rule of probability Independence of the conceptualization Ex. Two IVs in ANOVA, error term Independence of the measurement procedure Non-intercorrelated b/w variable Sequential analysis Dealing with the shared variance Combining variable Linear combination of variables Multivariate Data format Data matrix Raw data Correlation matrix (R) Standardized associaton among variables Variance-covariance matrix (∑) Averaged variation of the measurement Sum-of-square and cross-products matr