验孕棒验孕.pptVIP

Exhibit 18-7 The amount of common variance in X and Y may be summarized by r2, the coefficient of determination. As shown in the exhibit in the slide, the overlap between the two variables is the proportion of their common or shared variance. The area of overlap represents the percentage of the total relationship accounted for by one variable or the other. In the slide, 86% of the variance in X is explained by Y and vice versa. 第九章 假設檢定與關聯性分析 A correlation coefficient of any magnitude or sign, regardless of its statistical significance, does not imply causation. Correlation provides no evidence of cause and effect. There are several alternate explanations that may be provided for correlation results and these are listed in the slide. Note that causation may actually exist, but the correlation only shows that a relationship exists. Therefore, x may cause y or y may cause x. However, it may also be that z causes x and y or that y and x interact. 第九章 假設檢定與關聯性分析 Exhibit 18-8 Exhibit 18-8 Artifact correlations occur where distinct groups combine to give the impression of one. Artifact correlations should be avoided because they appear as one group when, in fact, there are distinct groups present. The left panel shows data from two business sectors. If all the data points for x and y variables are aggregated and a correlation is computed for a single group, a positive correlation results. Separate calculations for each sector reveal no relationship. In the right panel, the companies in the financial sector score high on assets and low in sales but they are all banks. When the data for banks are removed, the correlation is nearly perfect. When banks are kept in, the overall relationship drops considerably. 第九章 假設檢定與關聯性分析 In many relationships, other factors combine to make the coefficient’s meaning misleading. For example, there is a relationship between the results of presidential elections and the results of the Washington Redskins’ game before the election.