加载中…在因子分析之前是不是必须要通过KMOKMO检验与Bartlett球形检验？-
(2010-11-08 06:46:15)
标签：
检验球形因子分析杂谈 |
分类： 日常 |
Yes. These two tests are often used to examine the appropriateness
of data for factor analysis performance. It is often suggested that
KMO should be of 0.5 as a minimal level. Bartlett test should be
significant.
However, these tests are only two common (but not the only two) tests to examine the appropriateness. Additional inspections could include, as suggested, researchers should consider the correlation coefficient matrix. I'd like to recommend you to check: 1) whether you have many coefficent values less than 0.3 - which means that the correlations between the variables are too weak to (meaningfully) factor, and
2) whether you have many coefficent values greater than 0.7 - which indicates that the variables are highly correated and in fact they are measuring the same aspect of a construct. This could refer to multicollinearity issues. After you inspect for the 'extreme' values of R², I suppose you would be a clearer position to answer your question:
However, these tests are only two common (but not the only two) tests to examine the appropriateness. Additional inspections could include, as suggested, researchers should consider the correlation coefficient matrix. I'd like to recommend you to check: 1) whether you have many coefficent values less than 0.3 - which means that the correlations between the variables are too weak to (meaningfully) factor, and
2) whether you have many coefficent values greater than 0.7 - which indicates that the variables are highly correated and in fact they are measuring the same aspect of a construct. This could refer to multicollinearity issues. After you inspect for the 'extreme' values of R², I suppose you would be a clearer position to answer your question:
喜欢
0
赠金笔