多分类有序logit模型资料平行线假设及检验方法

The Parallel Line Assumption of Ordinal Logit Regression Model and its Test

目的介绍多分类有序logit模型资料平行线假设及检验方法。方法对儿童脂肪水平数据的自变量年龄、性别、胸围和瘦素等采用得分检验和Wald检验,以检查是否满足平行线假设。结果得分检验和Wald检验表明胸围和瘦素在不同的脂肪水平间不满足平行线假设。结论一般多分类有序logit模型中,自变量满足平行线假设,在违背假设时需要选用其他更合适的统计模型。

Objective To introduce the parallel line assumption of ordinal logit regression model and its test. Methods In order to example if the data of children fat level met the parallel line assumption, Applied the score test and wald test to the independent variables including age,gender,chest circumference and leptin. Results The score and wald test indicated that the age and lepin violated the assumption. Conclusion Under general ordinal logit regression model,if the parallel line assumption was violated,it should choose the appropriate statistic model.