请问：如何在有序逻辑回归（ordinal regression）中检验变量间的多重共线性？

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

请问：如何在有序逻辑回归（ordinal regression）中检验变量间的多重共线性？我在分析是遇到这样的问题，求高手解答。

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：regression regressio regress ordinal 有序逻辑回归 检验 如何

这个我也不懂啊

自己顶下

简单，你可以直接将结果变量y视为连续变量，然后进行OLS回归，不就知道自变量之间是否会出现多重共线性问题了么。注意，这样用的keypoint在于只需要知道自变量之间的多重共线性程度，而回归系数可以暂时不管。

喔，谢谢了，我这样做了以后，发现真的没有共线性

请问LZ做模型诊断怎么进行残差分析的？

By definition, multicollinearity arises when the predictor variables are strong correlated among themselves. In such a case, multicollniearity inflates the errors. Why don't you examine the correlation matrix of predictor variables if they are measured in continuous scales and see whether their correlation coefficients are way too high. The problem of multicollniearity might arises when r is greater than .80 (this is a rule of thumb, just for guidance). Because parsimony is praised in reductionist science, the number of predictors can be reduced if they are substantially correlated among themselves.

Greetings!
- You can use the linear regression procedure for this purpose. Multicollinearity statistics in regression concern the relationships among the predictors, ignoring the dependent variable. So, you can run REGRESSION with the same list of predictors and dependent variable.
- If you have categorical predictors in your model, you will need to transform these to sets of dummy variables to run collinearity analysis in REGRESSION.

Before analyzing any set of variables in a linear model, including ordinal regression, begin by check for multicollinearity by using linear regression to check the model Y= B0 + B1X1 + B2X2 + B3X + ... + e.
In SAS, this can be done with Proc Reg with the /VIF option on the model statement. Even if Y is binary, the linear regression procedure can be used to check for multicollinearity. Just use the linear regression procedure to screen for multicollinearity. Do your actual analysis with a logistic regression procedure.

The general guideline is that VIF values under 10 indicate that multicollinearity is not a problem. If you get high VIF values, then you will have to re-code your variables, do principle components analysis, or drop a variable.

Run a principal components analysis in SPSS on your independent variables. One of the options should be to use multiple R-square for the initial communalities. SPSS will then print those out. Since 1 – R-square is the tolerance, very large communalities indicate multicollinearity.