Berghaus Roy D, Lombard Jason E, Gardner Ian A, Farver Thomas B
Department of Population Health and Reproduction, School of Veterinary Medicine, University of California-Davis, Davis, CA 95616, USA.
Prev Vet Med. 2005 Dec 12;72(3-4):291-309. doi: 10.1016/j.prevetmed.2005.07.011. Epub 2005 Sep 1.
Factor analysis was used to examine the interrelationships among 38 variables collected as part of a Johne's disease risk assessment questionnaire completed in 2002 on 815 U.S. dairy operations. Eleven factors were extracted, accounting for two-thirds of the variance encountered in the original variables. Responses to many of the risk assessment questions were closely related. Standardized scores on the 11 factors were calculated for operations providing complete information, and were evaluated as predictors in a model-based logistic regression analysis with the outcome being whether operations had observed one or more cows with clinical signs suggestive of paratuberculosis during the previous year. A logistic regression model was also used to evaluate the predictive ability of a reduced subset of approximately one-third of the original variables that was selected to represent the derived factors. The performance of both sets of predictors was comparable with respect to goodness-of-fit and predictive ability. In conclusion, the length of the current risk assessment instrument could be reduced considerably without a substantial loss of information by removing or combining questions that are strongly correlated.
因子分析用于检验2002年对815家美国奶牛场进行的约翰氏病风险评估问卷调查中收集的38个变量之间的相互关系。提取了11个因子，占原始变量中所遇到方差的三分之二。对许多风险评估问题的回答密切相关。为提供完整信息的奶牛场计算了11个因子的标准化分数，并在基于模型的逻辑回归分析中作为预测因子进行评估，结果是这些奶牛场在前一年是否观察到一头或多头有副结核病临床症状的奶牛。还使用逻辑回归模型评估了从原始变量中选出的约三分之一的简化子集的预测能力，这些变量被选来代表派生因子。两组预测因子在拟合优度和预测能力方面表现相当。总之，通过去除或合并高度相关的问题，当前风险评估工具的长度可以大幅缩短，而不会大量丢失信息。
