Question

我使用svyglm运行以下逻辑回归模型和调查数据：

model7 <- svyglm(hivcode~I(age)+I(age)*factor(gendercode)+factor(gendercode)+
                 factor(maritalstatuscode)+factor(literacycode)+
                 I(age)*factor(maritalstatuscode)+
                 factor(residencecode),design=shs,family=quasibinomial)

我试图使用（像这样的）获得预测的概率：

newdata1<-with(new.clean,data.frame(age=mean(age),gendercode=factor(1),
                                    maritalstatuscode=factor(1),
                                    literacycode=factor(1),residencecode=factor(1)))

和

newdataP <- predeict(model7,newdata=newdata1,type="response").

首先我收到以下错误消息：

Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) :

我还想知道如何指定具体的因素水平，例如男性＆＃39;对于性别，＆＃39;单身＆＃39;对于预测代码中的婚姻状况等。

Answer 1

我需要更多关于要评论的错误的信息（可重现的示例效果最好），但对于性别和婚姻状况，请在原始数据shs中定义这些类（例如levels(shs$gender) <- c("male", "female") - 性别的顺序应该适应你的实际订单，我怀疑它目前被编码为数字？）。

要根据性别或婚姻状况进行预测，只需构建一个适当的数据框架，其中应包含模型所具有的所有变量。函数expand.grid可以创造奇迹。

expand.grid(gender = c("male", "female"), maritalstatus = c("married", "single"), numberofchildren = 0:5)
   gender maritalstatus numberofchildren
1    male       married                0
2  female       married                0
3    male        single                0
4  female        single                0
5    male       married                1
6  female       married                1
7    male        single                1
8  female        single                1
9    male       married                2
10 female       married                2
11   male        single                2
12 female        single                2
13   male       married                3
14 female       married                3
15   male        single                3
16 female        single                3
17   male       married                4
18 female       married                4
19   male        single                4
20 female        single                4
21   male       married                5
22 female       married                5
23   male        single                5
24 female        single                5

在逻辑回归模型中预测概率

1 个答案: