raushon 现在用multinom()拟合了一个模型 然后计算了logistic 回归系数的 relative risk ratio值 在医学界就是 “相对危险比率”的概念 , R中计算的时候就是给回归系数按e取幂 RRR_j = exp(beta_j) 也就是logistic 回归中的 Odds Ratio 值 目前需要依据RRR值来解释预测变量 和 响应变量 之间的定量关系 大家有没有做这方面研究的同仁,能否提供一点点相关的好文章查阅一下 谢了!
yihui 正好前一段时间看了一篇文档应该适合你的问题 Interpreting logistic regression in all its forms: http://www.ats.ucla.edu/stat/stata/library/sg124.pdf
raushon Is there anyone here could help me to interpret the following formula: P(y=1|x+1) / P(y=base category|x=1) Relative Risk Ratio = ------------------------------------------------ P(y=1|x) / P(y=base category|x) Especially, what's the meaning of " base category " in multinomal logistic regression Is it the fiducial class when model is fitted ? Heared about that Relative Risk Ratio is an concept in the field of epidemiologic statistics and also the concept of " base category " Look forward to your replay, Thank you !