我在我的模型中有这些预测变量，包括我使用“：”定义的2个交互项
> c(predictors)
> [1] "factor(Satisfaction)" "Gender" "Function"
> [4] "Gender:factor(Satisfaction)" "Gender:Function"
我想通过使用GLM中的terms语句来冻结变量的顺序，从而也冻结交互变量，我使用：
> fit <- glm(terms(form, keep.order = TRUE), data = data1, family = binomial)
但这是结果呢？第一个互动术语中的主要影响，即性别，首先没有报道？这也是我不使用条款语句时发生的情况（结果未显示） . 我认为条款声明可以解决问题，但可能不是？
> summary(fit)
> Call:
> glm(formula = terms(form, keep.order = TRUE), family = binomial,
> data = data1)
> Deviance Residuals:
Min 1Q Median 3Q Max
-1.7524 -1.0510 -0.5916 1.1169 1.9128
> Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 1.16293 1.13709 1.023 0.3064
factor(Satisfaction)2 -0.76457 0.96568 -0.792 0.4285
factor(Satisfaction)3 -1.09794 1.03372 -1.062 0.2882
Gender 0.90190 1.35310 0.667 0.5051
Function -0.02726 0.08460 -0.322 0.7473
factor(Satisfaction)2:Gender 0.32617 1.07892 0.302 0.7624
factor(Satisfaction)3:Gender 0.46506 1.18586 0.392 0.6949
Gender:Function -0.16573 0.09875 -1.678 0.0933
但是，当我在预测变量列表中使用Gender时，也可以不使用术语声明 . 这里发生了什么，我怎么能确定主效应总是首先报告与分类变量的交互？
谢谢你的时间和反应！
> predictors
[1] "Gender" "factor(Satisfaction)" "Function"
[4] "Gender:factor(Satisfaction)" "Gender:Function"
> print(form)
Chronic ~ Gender + factor(Satisfaction) + Function + Gender:factor(Satisfaction) + Gender:Function
> fit <- glm(form, data = data1, family = binomial)
Call:
glm(formula = form, family = binomial, data = data1)
Deviance Residuals:
Min 1Q Median 3Q Max
-1.7524 -1.0510 -0.5916 1.1169 1.9128
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 1.16293 1.13709 1.023 0.3064
Gender 0.90190 1.35310 0.667 0.5051
factor(Satisfaction)2 -0.76457 0.96568 -0.792 0.4285
factor(Satisfaction)3 -1.09794 1.03372 -1.062 0.2882
Function -0.02726 0.08460 -0.322 0.7473
Gender:factor(Satisfaction)2 0.32617 1.07892 0.302 0.7624
Gender:factor(Satisfaction)3 0.46506 1.18586 0.392 0.6949
Gender:Function -0.16573 0.09875 -1.678 0.0933 .