Predict mean response for a logistic regression model in R-CodePudding

I trained a logistic regression model in R using the glm function

model<-glm(df1$deny~df1$dir df1$hir df1$lvr df1$ccs df1$mcs df1$pbcr df1$dmi df1$self df1$single df1$uria df1$condominium df1$black,data=df1,family='binomial')

Now i want to get the mean response for a data point

test<-c(0.59,0.24,0.941177,3,2,0,1,0,0,10.6,1,1)

the test data points are the respective predictors as in the model. i.e. dir = 0.59, hir = 0.24...

How to obtain the mean response in this case?

CodePudding user response：

model <- glm(deny~dir hir lvr ccs mcs pbcr dmi 
                self single uria condominium black,
              data=df1,family='binomial')
test <- c(0.59,0.24,0.941177,3,2,0,1,0,0,10.6,1,1)

You can either use the model definition:

X <- matrix(c(1, test), nrow = 1)
beta <- coef(model)
drop(plogis(X %*% beta))

dftest <- as.data.frame(X)
names(dftest) <- c("dir", "hir", "lvr", "ccs", ...)

(you need to complete the list of names yourself, I'm lazy)

or possibly

names(dftest) <- setdiff(names(df1), "deny")

if the model variables match the order etc. of the data frame

Then:

predict(model, newdata = dftest, type = "response")

CodePudding user response：

Sorted. I did

df.test<- df1[0,-13]
head(df.test)
test<- c(0.59,0.24,0.941177,3,2,0,1,0,0,10.6,1,1)
df.test[nrow(df.test) 1,]=test
pred<- model.1 %>% predict(df.test,type='response')
pred